Towards understanding two-level-systems in amorphous solids - Insights from quantum circuits

Clemens M¨uller,1, 2, 3, ∗ Jared H. Cole,4, † and J¨urgenLisenfeld5, ‡ 1IBM Research Zurich, 8803 R¨uschlikon,Switzerland 2Institute for Theoretical Physics, ETH Z¨urich,8093 Z¨urich,Switzerland 3ARC Centre of Excellence for Engineered Quantum Systems, School of Mathematics and Physics, The University of Queensland, Brisbane, Queensland 4072, Australia 4Chemical and Quantum Physics, School of Science, RMIT University, Melbourne, Victoria 3001, Australia 5Physikalisches Institut, Karlsruhe Institute of Technology, 76131 Karlsruhe, Germany (Dated: October 11, 2019) Amorphous solids show surprisingly universal behaviour at low temperatures. The prevailing wisdom is that this can be explained by the existence of two-state defects within the material. The so-called standard tunneling model has become the established framework to explain these results, yet it still leaves the central question essentially unanswered - what are these two-level defects? This question has recently taken on a new urgency with the rise of superconducting circuits in quantum computing, circuit quantum electrodynamics, magnetometry, electrometry and metrology. Superconducting circuits made from aluminium or niobium are fundamentally limited by losses due to two-level defects within the amorphous oxide layers encasing them. On the other hand, these circuits also provide a novel and effective method for studying the very defects which limit their operation. We can now go beyond ensemble measurements and probe individual defects - observing the quantum nature of their dynamics and studying their formation, their behaviour as a function of applied field, strain, temperature and other properties. This article reviews the plethora of recent experimental results in this area and discusses the various theoretical models which have been used to describe the observations. In doing so, it summarises the current approaches to solving this fundamentally important problem in solid-state physics.

PACS numbers: 85.25.-j,74.50.+r,66.35.+a,63.50.Lm, 85.25.Cp

CONTENTS A. TLS microwave spectroscopy 13 B. TLS strain-spectroscopy 14 I. Introduction2 C. Quantum dynamics of individual TLS 15 D. Dielectric loss and participation ratio 16 II. Background4 A. Standard tunneling model4 VI. Experiments with superconducting resonators 16 B. TLS as a source of low-frequency noise5 A. Power-dependent dielectric loss 17 C. TLS as source of high-frequency noise6 B. TLS induced resonance frequency shift 18 C. Noise generated by TLS 18 III. Proposed microscopic models for the origin of D. TLS density measurements 19 TLS6 E. Electric field tuning of TLS 20 A. Tunneling atoms6 B. Tunneling electrons7 VII. TLS in other devices 20 C. Spins and magnetic impurities8 D. Emergent models8 VIII. Emergence of TLS in fabrication 21 A. Removal of dielectrics 21 IV. TLS interactions9 B. Superconducting materials and film A. Interactions with quantum devices9 deposition 21 B. Interactions with their dissipative C. Dielectrics 22 arXiv:1705.01108v3 [cond-mat.mes-hall] 10 Oct 2019 environment 11 D. Substrates 22 C. TLS-TLS interactions 11 E. Cleaning methods, chemical residuals and film structuring 22 V. Experiments with quantum bits 13 F. Epitaxial films 23

IX. Conclusions and outlook 24

∗ [email protected] Acknowledgments 25 † [email protected] ‡ [email protected] Table of symbols 25 2

References 25 ence and parameter fluctuations for superconducting qubits [11]. At present, quantum circuits are typically fabricated from superconducting aluminium because it I. INTRODUCTION allows the formation of high-quality JJs using the well- established techniques of double angle shadow evapora- Two-Level-Systems in amorphous materials - tion [12–14] (see Fig.1a) for an example). Once the sam- The properties of amorphous or glassy material are dom- ple chips are exposed to air, an amorphous oxide layer inated by the absence of long-range order in the atomic will grow on any exposed aluminium structure, which lattice. Even though they have been studied for decades, is characterized by a large dielectric loss that in turn is there is still a surprisingly large amount that is not un- attributed to high TLS densities in the amorphous ma- derstood about such materials. Despite the randomness terial [15]. Moreover, the insulating tunnel barrier of JJs of the atomic arrangements, and even independent of is itself also made from amorphous aluminium oxide and their chemical composition, most amorphous solids dis- thought to host TLS. Figure1b) illustrates some models play surprising similarities in their properties at temper- of TLS formation in Josephson junctions. atures below a few Kelvin. This universality is typically When the two states of a TLS are associated with the explained via the so-called standard tunneling model displacement of a charge, they possess an electric dipole (STM), whose basic principle is that the low-temperature moment which couples them to the oscillating electric behaviour of glassy systems is dominated by the presence field present in capacitive circuit components. For TLS of two-level defects (TLS) within the material. These de- residing in the typically few nm-thin tunnel barrier of fects are in general not due to impurities inside the ma- JJs, where the electric field strength can reach several terials, but rather emerge from the deviations away from 100 V/m, this coupling may become particularly strong. crystalline order which characterise the amorphous state. While such strongly coupled TLS are especially detrimen- Due to their low energy, such two-level defects are typ- tal to qubit operation, their coherent interaction with the ically saturated at high temperatures. However as the circuit dynamics provides a pathway towards direct ma- material is cooled, these additional degrees of freedom nipulation and readout of the state of the microscopic become available and can dominate the low temperature defects using their macroscopic host device. properties. When we consider TLS in quantum devices, there In general, the STM is exceptionally good at describ- is a useful classification to bear in mind which relates ing the low-temperature behaviour of most amorphous to whether the TLS internal dynamics on experimental materials [1,2]. As this approach treats the defects at timescales is dominated by incoherent or processes or not. a phenomenological level, it can be used across many Each individual TLS is coupled to an environment at am- different systems. However, it leaves one fundamentally bient temperature T which might for example consist of important question unanswered - what is the underlying phonon modes, other TLS in their vicinity or quasipar- microscopic nature of the two-level defects? Two-Level-Systems in Quantum Devices - Re- a) c) cently, TLS have attracted substantial renewed interest because they are seen as a major source of noise and decoherence in superconducting quantum devices. This d) includes superconducting quantum bits (qubits), which are circuits with resonance frequencies in the microwave 400 nm 1 mm range, tailored from microstructured inductors, capaci- b) Al tors, and Josephson tunnel junctions (JJs) [3–5]. To op- e- erate such a circuit as a qubit, it is necessary to have two x long-living eigenstates which are used as logical states |0i

and |1i, and between which transitions can be driven to AlO ~ 2 nm realize logical quantum gates. Since such circuits in gen- Al eral have more than two excited states, they need to be sufficiently anharmonic so that all transition frequencies Figure 1. a) SEM photograph of a Josephson junction (JJ) are unique and can be separately addressed. This is re- made with the aluminium shadow evaporation technique. b) Sketch of a JJ formed by two superconducting Al electrodes alized by incorporating Josephson junctions, which can that are separated by a thin (≈ 2 − 3 nm) layer of amorphous be modelled as nonlinear inductors whose value is tuned AlOx dielectric, here illustrated as hosting TLS formed by via a bias current or an applied magnetic flux. Since tunneling atoms, dangling bonds, and trapped charges. c) the first observation of coherent quantum dynamics in Photograph of a 3D-Transmon qubit, showing the opened a Cooper pair box in the year 1999 [8], superconducting 3-dimensional cavity and a qubit chip in its centre, similar qubits have evolved into one of the leading contenders for to Paik et al. [6]. d) Planar Transmon qubits, consisting of the realization of large-scale quantum computing [9, 10]. cross-shaped capacitor electrodes shunted by JJs. Meander However, loss and fluctuations due to parasitic cou- structures are resonators coupled to a transmission line for pling to TLS present a significant source of decoher- qubit readout, similar to Barends et al. [7]. 3 ticles formed from residual non-superconducting charge dynamics is overdamped. carriers. This coupling leads to incoherent transitions be- State of the art - The impressive enhancement of tween the TLS eigenstates (dissipation and excitation) as coherence times that was seen in experiments with super- well as random fluctuations in their energy (dephasing). conducting qubits during the last decade [5, 22], was to Fluctuators can be defined as TLS who are in strong a large extend achieved by reducing the coupling to TLS contact with their own environment and incoherently flip (and other decoherence sources) with improved circuit between two states on typical experimental timescales. designs, rather than by lowering the TLS densities by im- These incoherent state transitions are due to a combi- proving materials and fabrication procedures. The best nation of quantum tunnelling through the barrier and performing qubits today employ small tunnel junctions decoherence due to the coupling to their environment. to reduce the number of strongly coupled TLS, and em- The additional option of thermal activation, i.e. when ploy circuit layouts in which the electric field concentra- the thermal energy kBT of the environment is larger tion is reduced at lossy interfaces, e.g., by increasing the than the height of the barrier between the two wells, distances between capacitor electrodes. This is clearly is typically not relevant for the tunneling two-level sys- demonstrated in so-called three-dimensional “Transmon” tems considered here. In low-temperature electronics the qubits [6, 23], which feature large capacitor plate sepa- fluctuators can couple to their host circuit, and the re- ration and are placed into cavity resonators having large sulting fluctuations will manifest themselves as quasi- volumes of a few cm3, such that the electric field strength classical random variations in circuit parameters. In is significantly lower than in other designs (see Fig.1c) addition, through defect-defect coupling they may also for an example). Careful revision of clean-room recipes modify the behaviour of other TLS in the ensemble. Since has been identified as a second necessity in order to avoid very fast fluctuations will average out over experimen- the formation of TLS during sample fabrication. tal timescales, fluctuators will typically modify the dy- The growing understanding of how to evade the TLS namics of quantum devices via contributions to the low- problem has brought superconducting circuits in short frequency environmental noise spectrum [16], which in time to the verge of becoming scaled up to integrated turn is mostly responsible for loss of phase coherence in quantum processors. Nevertheless, dielectric loss remains these devices [17]. responsible for the major part of energy relaxation in Coherent TLS, on the other hand, are those where state-of-the art qubits [24], and coupling to even sparse the coupling between the TLS and their environment is TLS baths causes relaxation [25] and dephasing [26]. weak enough that they can remain in one of their eigen- These issues will gain in importance once circuits com- states, or they can even be placed in a coherent super- prise more than a handful of prototype qubits, and thus position of states, over the timescale of an experiment. must be urgently addressed to ensure continuation on the Typically, coherent TLS will have energy splittings that path towards a solid-state quantum computer. are larger than the thermal energy of their environment, On the other hand is it just this sensitivity to even E > kBT , such that incoherent excitations into their ex- single defects that makes superconducting circuits ideal cited state are suppressed and their equilibrium steady- tools for the study of TLS and material dissipation state will be their ground state. Such coherent TLS can mechanisms in the quantum (i.e. single-photon and low- even reach the strong coupling regime with their host cir- temperature) regime. The possibility of using supercon- cuit or each other, which is characterized by a coupling ducting qubits to probe individual defects has fundamen- strength that exceeds the decoherence rates of both the tally changed both the questions that can be asked and TLS and its hosting device. This strong coupling results that need to be asked. These circuits allow not just the in modifications of the energy level structure and quan- dissipative dynamics of a defect to be resolved directly, tum dynamics of the hosting device which can be directly but even enable one to measure and manipulate the quan- observed, for example as anti-crossings in qubit spec- tum states of coherent TLS. This has opened doors to troscopy [18] or coherent beating in population dynam- new tests and studies of the nature of individual defects ics [19]. Coherent TLS in their groundstate are also able and therefore the ensemble as a whole. to resonantly absorb energy from their host quantum cir- Outline - In this review, we focus on TLS which cuits and dissipate it into their own environment [7, 20]. affect novel superconducting quantum circuits such as The distinction between fluctuators and coherent TLS qubits and resonators. With qubits, one is able to ac- introduced here is not fundamental and they are in fact cess and control the quantum states of individual TLS, thought to be formed by the same physical entities. How- enabling novel studies of their mutual interactions and ever it provides a useful distinction when considering the decoherence mechanisms. Microwave resonators on the possible dynamical effects that can arise from an individ- other hand are effective tools to characterize loss from ual TLS or the effect of an entire ensemble [21]. defects at layer interfaces and to quickly validate fab- We also note that the classification given above is dis- rication processes. They are also a necessary part of tinct from the term “incoherent TLS”, which in the glass the leading solid-state architecture for quantum compu- physics community is taken to imply a TLS for which in- tation, circuit-QED [27, 28], where qubits are coupled to coherent, environment induced processes dominate over resonators to improve interactions and readout. We fur- all other energy scales, or in other words a TLS whose ther review the existing theoretical models for the origin 4 of TLS and how these can be reconciled with existing and [37], and Lubchenko and Wolynes [38]. future experiments. The standard tunneling model makes the following key The review begins with a short introduction of the assumptions standard tunneling model, and how ensembles of TLS are a source of low- and high-frequency noise for super- • The TLS can exist in one of two energetically sim- conducting circuits. We continue in Sec.III with an ilar configurations. overview of the plethora of proposed microscopic mod- els for the origin of TLS. The following SectionIV gives • These configurations are modelled as two minima a brief discussion of the basic physical mechanisms by in a double-well potential which are separated by a which TLS interact with superconducting circuits and barrier. their environment and how these interactions can be utilised to draw conclusions on their microscopic origin. • At sufficiently low temperatures, thermal activa- SectionV presents an overview of experiments on qubits tion over the barrier is suppressed and the dynam- such as spectroscopy, by which the presence of coherent ics are governed by quantum tunneling through the coupling to individual TLS was first revealed. It also barrier. reviews measurements of the coherent time evolution of • In general, the system couples to applied electric or TLS quantum states, including quantum state swapping, strain fields in such a way that transitions can be creation of TLS entanglement, measurements of TLS de- driven between the states. coherence, and studies of mutual TLS-TLS interaction. Most of the knowledge about materials and fabrication • Due to the random atomic arrangements, an en- steps which give rise to TLS formation has been obtained semble of TLS is characterized by a wide distribu- from experiments on superconducting resonators, which tion of potential barrier heights and thus spans a we describe in Sec.VI. The following Sec.VII gives a large range of switching rates and eigenenergies. very brief summary of other experimental architectures where TLS are believed to be of relevance, such as nano- A common visualisation of a TLS in the STM is given mechanical resonators. Finally, we give an overview of in Fig.2 where a particle can sit in one of two parabolic recent progress and results from fabrication of supercon- potential wells. The energy asymmetry of the ground- ducting circuits in Sec.VIII, which also includes a re- state wavefunctions of the two wells is labelled ε. This view of ongoing efforts to fabricate Josephson junction asymmetry can be due to differences in width or shape with crystalline tunnel barriers. For additional recent of the two wells, or of the classical minimum energy of reviews on the importance of materials in superconduct- the wells (as depicted). The energy associated with the ing quantum bits we refer the reader also to McDermott process of tunneling through the barrier separating the [29] and Oliver and Welander [30]. We end with a short two wells is ∆0. summary and outlook, aiming to provide a perspective The effective Hamiltonian for this situation has the for future experimental and theoretical efforts to deter- form mine the microscopic origins of TLS.   1 ε ∆0 1 (p) 1 (p) HTLS = = εσz + ∆0σx , (1) 2 ∆0 −ε 2 2 II. BACKGROUND (p) with the Pauli matrices in the position basis σz = A. Standard tunneling model (p) |RihR| − |LihL| and σx = |RihL| + |LihR|. Note that for clarity, our notation here differs from the usual STM The concept that a single atom can tunnel between literature where typically the asymmetry energy ε is la- energetically equivalent local potential wells was under- belled ∆. Due to the tunneling between the wells, the stood even at the birth of quantum mechanics [31, 32]. two lowest eigenstates in the left (|Li) and right (|Ri) Soon after, Pauling [33] discussed the implications for the well hybridise and form the eigenstates oscillatory and rotational motion of atoms in molecules θ  θ  and crystals. However, more recently there has been a |ψ i = sin |Li + cos |Ri , (2) vast array of different models proposed for the physi- + 2 2 cal origin of TLS. The details of many of these mod- θ  θ  |ψ i = cos |Li − sin |Ri , (3) els will be discussed in sectionIII but they all share − 2 2 some fundamental properties which form the basic com- ponents of the standard tunneling model (STM) [34– where the mixing angle θ is defined via tan θ = ∆0/ε. We 36]. In this section, we briefly summarise the key as- can then rewrite the Hamiltonian in the basis of eigen- pects of the standard tunneling model which will be states as important for the later discussions. A more in-depth 1 overview of the STM and its supporting experiments is H = Eσ , (4) given in Enss and Hunklinger [1], Esquinazi [2], W¨urger TLS 2 z 5 with the energy difference between the eigenstates which is typically approximately constant over the en- ergy ranges of interest. Here ∆min refers to the mini- q 0 2 2 mum tunneling energy, below which the particle can no E = E+ − E− = ε + ∆ . (5) 0 longer tunnel between the wells and the TLS is essentially just two local minima in a classical sense. Typically, the In the limit |ε|  ∆ , the eigenstates are well described 0 observation timescale of an experiment (which can vary by the left and right well states. However for |ε| ≈ 0, the from 10−6 to 103 seconds) sets a minimum energy scale eigenstates are a superposition of the two well states. of interest. In amorphous glasses in general, the distri- bution of ∆0 extends far below any energy set by the min experimental timescale (i.e. ∆0  E) which supports the constant density of states approximation. Using this density of states and computing the specific heat of a

energy material containing TLS defects, one finds that CV ∝ T , as opposed to the usual T 3 obtained from Debye theory. This modified temperature dependence is one of several key predictions of the STM for the low-temperature be- haviour of amorphous solids. Although the STM typically assumes independent position TLS, including the effects of interactions between TLS leads to corrections to the low-temperature response of Figure 2. Double-well potential modelling a TLS. The en- glasses due to the formation of collective states. Al- ergy difference E between the TLS eigenstates |ψ+i and |ψ−i though such extensions to the STM have been studied in is determined by the asymmetry energy ε and the inter-well depth in glasses both experimentally and theoretically, tunneling rate ∆ . 0 recent work with superconducting circuits has provided far more direct evidence for TLS-TLS interactions. We Using Wentzel-Kramers-Brillouin (WKB) theory, we elaborate more on such models in sectionIIID where we can estimate the value of ∆0 in terms of the barrier height provide examples of collective models for TLS origins, V , the spacing between the wells d and the effective mass and sectionIV, where we discuss direct probes of TLS- of the particle m [2], giving TLS interactions.

−λ ∆0 = ~ω0e , (6) B. TLS as a source of low-frequency noise where r 2mV The STM and its extensions are often used as models of λ = 2 d. (7) electrical noise in general, especially in low-temperature ~ electronic devices. Typically, this noise is thought to arise

The scale factor ω0 depends on the exact functional form as an ensemble effect from a large number of TLS with of the potential, see Philips [36] for examples. However, some distribution of energies and tunnelling rates. Here, this detail is typically unimportant for the overall be- we briefly touch on noise in electrical circuits in general, haviour of the system given the exponential dependence so as to introduce the concepts and notation relevant on λ. to our later discussion. We will focus first on the low- In contrast to disorder or impurity defects in crys- frequency noise spectrum, and discuss the high-frequency tals which can also display TLS behaviour, see [1] for noise components relevant to energy dissipation in the a detailed discussion), the TLS parameters in amor- following section. phous solids vary from defect to defect due to the ran- When characterizing a fluctuating quantity X(t) (e.g. domised nature of the local atomic configurations. The an applied voltage), one usually defines the spectral func- STM assumes that the asymmetry ε and tunneling pa- tion of this quantity through the Fourier transformation rameter λ are independent and uniformly distributed, of its two-time correlation function P (ε, λ)dεdλ = P0dεdλ, where P0 is a constant. Re- Z ∞ −iωt expressing in terms of E and ∆0, SX (ω) = e hX(t)X(0)i . (10) −∞ E P (E, ∆0)d∆0dE = P0 d∆0dE , (8) At low frequencies and temperatures, the dominant noise p 2 2 ∆0 E − ∆0 source in electrical solid-state devices typically has the functional form and integrating over ∆0 gives the density of states, α SX (ω) ∝ ω , (11) Z E 2E D(E) = P (E, ∆0)d∆0 = P0 ln min ≈ D0 , (9) min ∆ which, when α ∼ −1, is the (in)famous ‘1/f noise’. ∆0 0 6

This contribution to the noise has been studied for the TLS’ dipole response. However, the recent work on decades and although it is ubiquitous across many differ- TLS physics in superconducting circuits has focused on ent types of devices and experiments, at this stage it is finding ways to not just understand, but actually re- still poorly understood. Recently, due to its detrimental move TLS within the metal oxide surfaces and junctions. effects on coherent superconducting circuits, the need to Therefore, a phenomenological model is insufficient, we understand the microscopic origin of this noise has been need to know what a TLS is, not just how it behaves. given a new urgency (see 39 and 17 for more specialised Although the double-well potential model illustrated in and in-depth reviews of 1/f noise in condensed matter Fig.2 provides an intuitive picture of how the TLS pa- physics and qubits). Qubit coherence depends on both rameters might come about, remaining ‘unknowns’ in the noise spectral function at the resonance frequency of the problem are microscopic parameters such as parti- the circuit and at or close to zero frequency. It is com- cle mass, the charge of the tunnelling entity, and the size mon that coherence times are limited by the 1/f noise and form of the TLS potential. power (see sectionV for more details) making the under- standing of its origins of paramount importance. The random nature and distribution of relaxation III. PROPOSED MICROSCOPIC MODELS FOR times characteristic of the STM already suggests that THE ORIGIN OF TLS an ensemble of TLS might provide a viable model for the 1/f noise. Although Dutta and Horn [16] showed To date, all reported experiments are broadly consis- several decades ago that a distribution of TLS switching tent with the assumption that the TLS in superconduct- rates P (γ) ∼ 1/γ together with the constant distribution ing circuits are equivalent to those known to exist in in energy from the STM results in the required spectral amorphous dielectrics such as glasses [35, 45]. While characteristics of the noise, the link between a specific these TLS in glasses have been studied intensively dur- case of a TLS ensemble and 1/f noise in the same device ing the last 40 years, their microscopic nature remains has so far been difficult to prove definitively. elusive [46, 47]. Many different proposals exist to explain the origin of TLS in amorphous oxides within superconducting cir- C. TLS as source of high-frequency noise cuits, the major categories of which are summarised in this section. Due to the random nature of the oxide struc- ture, there is no clear reason to expect unique spectral At high frequencies (~ω  kBT ) electrical noise typically scales with α > 0, where α ∼ 1 is referred to signatures of a particular TLS type - in contrast to typi- as “ohmic” noise. This is the Johnson-Nyquist limit cal defects in crystals which reside in a more well-defined of noise and is traditionally explained as originating environment. However, given the additional information from a collection of linear harmonic oscillators, i.e. the obtained from more recent experiments on strongly cou- modes of the waveguides and cables in the experiments pled (coherent) TLS, there is hope that an accurate com- and control electronics. However, Shnirman et al. [20] parison to theoretical models is possible. These prospects showed that assuming a distribution of STM parameters, have encouraged several groups to investigate detailed s computational models of TLS in amorphous material and P (ε, ∆0) = (ε/∆0) with −1 ≤ s ≤ 1, leads to a linear distribution of TLS energies E which results in a noise then attempt to estimate the values of the resulting TLS parameters and their response to pulse sequences, applied spectral density S ∝ 1/f for hf  kBT while at the same time giving the standard ‘Ohmic’ signature (S ∝ f) strain, electric fields or temperature variation. at high frequencies hf  kBT . Apart from ensembles, even sparse distributions of TLS with eigenenergies comparable to the circuit frequencies can lead to dis- A. Tunneling atoms sipation, as they can accept energy from the circuit and dissipate it into their own environment [40]. This One of the most physically appealing models is to as- will typically manifest in a noise spectral function with sume that the two-level system is formed by the literal pronounced peaks at certain frequencies, as observed movement of an atom or small group of atoms between commonly in superconducting qubits [6,7, 41, 42]. two potential minima. An illustration of such systems in The cross-over between low and high-frequency noise an amorphous material is shown in Fig.3, depicting the regimes has also been studied experimentally (see for motional degrees of freedom of atoms, dangling electronic example 43 and 44) and shown to appear at energies bonds, and hydrogen atoms. corresponding to the experimental temperature, con- Although this has traditionally been the physical pic- sistent with the theoretical model of Shnirman et al. [20] ture that is most often quoted for the STM, it still leaves open the question of which atom or degree of freedom is Summary - The utility of the STM introduced here actually ‘moving’. is that the majority of observed phenomena can be de- For accurate computational simulations of these mod- scribed with only a few parameters, namely ε, ∆0, the els one has to grapple with several problems from a quan- distributions of these parameters for a TLS ensemble and tum chemistry point of view. The energy splittings of 7

ergy pathway using the nudged elastic band method [52]. In particular, they showed that the formation of hydro- tunneling genated Al vacancies is energetically favourable and that atoms collective these defects form threefold degenerate rotors with tun- motion nel splittings in the MHz to GHz range. These defects also displayed electric dipole moments of approximately dangling 0.6 eA,˚ in the range typically observed in experiments. bonds Using empirical potentials, DuBois et al. [53] solved the Schr¨odingerequation for the position of the oxygen atom - in analogy with oxygen interstitial defects in crys- Hydrogen talline silicon and germanium. Varying the position of rotors surrounding aluminium atoms allows one to test vari- ous double-well configurations and compute the splittings Figure 3. Illustration of some example mechanisms of TLS to high precision using conventional finite-element tech- formation in an amorphous material: tunneling of single niques. Although one can find many atomic configura- atoms and collective motion of small atomic groups, dangling tions that show the correct range of splittings and charge bonds, and Hydrogen defects. dipoles, this still leaves the question of which atomic con- figuration is the correct one largely unanswered. There are also significant differences in energy scale depend- typical strongly coupled TLS observed in experiments ing on whether a one-, two- or three-dimensional model are 0.5 - 10 GHz, (ie. neV) which is far too small for is employed [53–55]. In principle, both these limitations the majority of ab-initio methods. In addition, the typi- can be addressed by taking realistic atomic positions from cal length scales of interest when studying metallic-oxides ab-initio molecular dynamics simulations [56]. However, are 1-100 nm, involving 100s, 1000s or even more atoms. at stoichiometries and densities which are representative Recent approaches to this problem have focussed on the of experimentally grown oxides, the resulting TLS have specific problem of aluminium-oxide tunnel junctions, splittings in the range of terahertz and above. This sug- most relevant for TLS in superconducting circuits, con- gests that the molecular environment within a junction trasting with the more generic glass studies of the 80s is too tightly constrained to permit appreciable delocal- and 90s. One is then able to use first principles methods, isation of a single oxygen atom [55], providing further or empirical and effective potentials which have been op- weight to the argument that clusters of atoms are in- timised to correctly model the aluminium-oxide bonding volved in forming the TLS [57]. configuration within the oxide. In order to explore TLS arising from more realistic One of the original suggestions for the origin of TLS atomic configurations for amorphous Al2O3, Paz et al. [58] performed molecular dynamics simulations at ‘el- in AlOx and SiO2 amorphous films observed with qubits are OH bonds or defects [48, 49]. Gordon et al. [50] dis- evated’ temperatures of 25K and searched for bistable cussed the possibility that hydrogen interstitials within switching between atomic configurations. The free en- ergy profile was then extracted for these configurations the Al2O3 lattice or at the surface could form suitable two-level defects. They considered the stability of the and barrier tunneling and charge dipoles estimated. This various charge states of the hydrogen interstitial and approach allowed the identification of general structural the resulting structural geometries, and then solved the motifs that display bistability without having to presup- Schr¨odingerequation for the hydrogen atom within the pose any particular symmetry of the defect. Several can- potential formed by the surrounding Al O structure - didate configurations were found with charge dipoles of 2 3 ˚ obtained directly from ab-initio methods. Although they order 0.9 eA and estimated energy splittings of 70 - 170 found electric dipole strengths commensurate with ex- GHz. However, the statistics of defect identification was perimental observations, the calculated tunnel splitting limited by both the amorphous nature of the structures as a function of O-O bond distance reached a minimum of and the complexity of the calculation. approximately 16 GHz. Such a lower limit is not seen in experiments and is inconsistent with the picture of TLS contributing to both low- and high-frequency noise (see B. Tunneling electrons sectionIIB). By a combination of ab-initio structures and single Following a similar philosophy to the tunneling atom body Schr¨odingerequations, Holder et al. [51] also in- models, the idea that single electrons can tunnel between vestigated the role hydrogen plays in hydrogenated Al local minima is also a very clear concept. Many of the vacancies, bulk hydrogen interstitial defects, and in a sur- earlier experiments on strongly coupled TLS where anal- face O-H rotor. In this work, the structure was also com- ysed in terms of such electron tunneling [59, 60] How- puted using ab-initio methods to determine defect forma- ever, this interpretation somewhat fell out of favour as tion energy and stability. The potential landscape seen the small energies of TLSs were taken as corroboration of by the rotor can then be extracted from the minimum en- tunneling atoms because of their larger mass as compared 8 to electrons. To contribute appreciably to thermal prop- limiting the sensitivity of a range of magnetometry and erties at sub-Kelvin temperatures, a significant number sensing applications as well as severely reducing coher- of TLSs need to have energies of the order of gigahertz, ence times of many superconducting qubit types [64, 65]. which was considered inconsistent with the typical eV A recent review by Paladino et al. [17] provides a com- energy scale for electrons in solids [61]. plete overview on this topic, and here we will only sum- More recently, the tunneling of electrons has been re- marise some more recent results to complete the picture. addressed considering more sophisticated effective mod- Experimental studies have shown that the magnetic els. Agarwal et al. [61] showed that if an electron mov- noise is predominantly generated in the native oxide en- ing between two wells is dressed by a collective phononic casing the circuits and initial measurements showed that state, this has the effect of renormalizing the effective it scales linearly with device dimensions [66, 67]. Assum- TLS parameters. This renormalization factor was esti- ing the noise to be generated by electron spins, densities mated to be of order e−10 ≈ 4.5 × 10−5 and the re- of ∼ 5 × 1017m−2 have been inferred from measurements sulting renormalised TLS parameters were found to be in a variety of SQUIDs [66] as well as on other, non- typical energy scales commensurate with experiments. superconducting material surfaces [68]. The decoherence channels and TLS-TLS interactions one Very recent work on spin states on the surface of metal- would expect for this model were also estimated and oxides has shown strong evidence of the role of oxygen are compatible with observations in superconducting cir- and spin-spin interactions. For example, Lee et al. in- cuits. Lastly, a detailed experimental approach to test vestigated surface vacancy states on Al2O3 and SiO2 this model was suggested, using phononic band-gap engi- using density functional theory [69, 70], and concluded neering of a metal-oxide junction to effectively suppress that dangling bond states on the surface can form para- the phonon dressing and thus gap out the TLS in the magnetic localised magnetic moments and can at least microwave range. Such an engineered device provides a partially explain the low frequency magnetic noise. In pathway to benchmarking different models and compar- following work, Wang et al. [71] used density functional ing their predictive power, even for amorphous devices theory to study molecular oxygen adsorbed to the surface with a distribution of parameters. of Al2O3, estimating a magnetic dipole of 1.8 µB. Us- Another effective single electron model is given ing Monte-Carlo simulations of a spin lattice, they show by localised metal-induced gap states (MIGS) at a that the flux noise generated by such a model was con- metal/insulator interface. In this case, disorder at the in- sistent with that observed in SQUIDs. This analysis terface localises a substantial fraction of MIGS electrons. corresponds well with (and was inspired by) recent ex- The TLS is then formed by the magnetic moment of this periments illustrating the role of molecular oxygen [72] localised electronic state. Choi et al. [62] performed a and hydrogen [73–75] in generating magnetic flux noise. tight-binding analysis of an exemplary metal/insulator It has been shown that careful surface treatments to re- interface and showed that the expected areal density move either of them results in significantly reduced loss in and resulting low-frequency noise spectral function are resonator circuits. It remains unclear what, if anything, consistent with observed data on magnetic field noise in of this analysis can explain the emergence of strong in- SQUIDs. Although this model was originally presented teraction between quantum circuits and TLS. As yet, no in an effort to explain the localised magnetic moments experiments have demonstrated a signature of adsorbed measured on the metal-oxide surface of SQUIDs, it may magnetic moments while also demonstrating strong co- equally apply to Josephson junction based defects - form- herent coupling of the same frequency-resolved entity to ing either a localised charge or spin defect. In this spirit, a qubit or resonator degree of freedom. the model was subsequently used to analyse the results of SET measurements of strongly coupled TLS [63] (see sec- tionVA). At this stage, little is understood on how sus- D. Emergent models ceptible MIGS are to decoherence and therefore whether the model is consistent with the long coherence times of There are various proposals of emergent TLS models, TLS observed in qubit and resonator experiments. in which the underlying degrees of freedom do not specif- ically display TLS-like behaviour but interaction with other collective degrees of freedom results in effective C. Spins and magnetic impurities TLS behavior. This general idea has also been studied in depth by the Another natural model for TLS is given by the intrin- glass community, for instance assuming localised phonon sic spin of electrons or atomic constituents, which may modes resulting from anharmonic local potential wells generate fluctuating magnetic moments. Magnetic noise (the ‘soft-phonon’ or ‘soft-potential’ model) see e.g. [76– in superconducting circuits has been studied extensively 85]. Although the STM works very well to explain the from both a theoretical and experimental point of view. thermal response of glasses below temperatures of 1K, The magnetic noise observed e.g. in superconducting distinct deviations from STM behaviour are observed be- quantum interference devices (SQUIDs) typically shows tween 1K and 10K, which could be explained in terms of − 1 a 1/f spectrum with an amplitude of A1/f ∼ 1µΦ0Hz 2 , such localised modes. More details on this model and 9 its predictions in glass physics can be found in chap- contrast, defects which are not inversion symmetric have ter 9 of Esquinazi [2]. However, as this model is largely higher frequencies (exceeding those typically probed indistinguishable from the STM at temperatures below in qubit or resonator experiments) and respond more 1K, it has received little experimental attention in su- strongly to phonons. This difference in phonon response perconducting devices. The rapid closing of the super- leads to three different TLS-TLS interaction energy conducting gap as system temperatures approach the scales via acoustic dipolar interactions: At sufficiently critical temperature of aluminium (Tc ≈ 1.2 K) makes low temperatures, the inversion asymmetric TLS ef- experimental tests in this regime extremely challenging. fectively ‘freeze-out’ due to their mutual interaction. Here instead we focus on models proposed recently in the Hereby, they form an effective irregular spin-lattice context of qubit and resonator experiments, or those in which imposes a disordered local strain field upon which metal oxides of interest to these experiments have the nominally symmetric low-frequency TLS. The been explicitly discussed. interaction energy scales therefore lead to a hierarchy One class of emergent model assumes pairs of trapping of different responses as a function of temperature levels in the oxide barrier coupled to the superconductor, and provide a plausible explanation for much of the where a Cooper-pair couples simultaneously to a pair of universal behaviour attributed to TLS. This model was states. This has been dubbed the Andreev-level fluctua- motivated and tested using disordered crystals [92–94] tor [86]. Population and depopulation of the trap levels but direct applicability to amorphous metal-oxides is yet provides the fluctuating charge coupling to the electric to be shown conclusively, although recent work suggests field in the dielectric oxide, and potentially also modifies that nonequilibruim absorption measurements provide a the critical current of the Josephson junction [87]. This method for probing such interacting TLS models [95, 96]. mechanism can provide the correct frequency and tem- perature behaviour for the noise generated by TLS, but Summary - The large number of microscopic models requires an unphysical high density of states for the elec- for TLS proposed in the literature poses a major chal- tron traps [88]. To remedy those shortcomings, Faoro and lenge when trying to identify clear candidates. Many Ioffe [88] suggest that strong on-site repulsion of trap lev- of the models are hard to distinguish experimentally, as els could lead to Kondo-like resonances close to the Fermi their signatures in the data are very similar. Experi- level. These resonances are characterized by a Kondo ments that probe several TLS properties simultaneously temperature TK , and the interplay between supercon- may be needed to finally unambiguously determine the ductivity and Kondo-physics determines the occupancy microscopic origin of the TLS, as we will discuss in the of the trap. For TK ∼ ∆ this mechanism leads to a high following. density of localized states at low energy, much larger than the original density of charge traps [88, 89]. Further, M¨uller et al. [40] conjectured that the ob- IV. TLS INTERACTIONS served strongly coupled TLS in phase qubits are formed from superradiant Dicke states of interacting microscopic Here we introduce the basic ideas of how TLS can cou- TLS, providing an explanation for the occurence of strong ple to the dynamical degrees of freedom of their environ- coupling for only a small number of TLS. In this model, ments, including superconducting circuits as well as each the strongly coupled TLS would exhibit higher levels other. These ideas are closely tied to the possible micro- with a quasi-linear level structure, an observation that scopic models for the origin of TLS as already reviewed is incompatible with experiments probing the structure in Sec.III, and careful analysis of these interactions may of those defects. lead to final identification of the TLS’ microscopic origin. Finally, also interactions between TLS can lead to strong modifications of their underlying properties and distributions. Coppersmith [90] conjectured that the uni- A. Interactions with quantum devices versality seen in the STM stems from very strong dipolar interactions between microscopic TLS. In this case frus- TLS in close vicinity to quantum circuits can couple tration of the interaction leads to some of the TLS being to their dynamics by three basic mechanisms, explained effectively decoupled from the rest of the ensemble and in detail below. An important concept here is the strong dynamically free. The properties of these free TLS will coupling regime between two quantum system, which in be universal in a large range of parameters [90]. this case can be reached when the coupling strength g Similarly, Schechter and Stamp [91] have proposed between the host circuit and an individual TLS is much that much of the universal behaviour seen in glasses larger then the dissipation rates of both circuit and TLS, comes about due to the interaction between two classes g > ΓQ, ΓTLS. Here ΓQ and ΓTLS are the decoherence of two-level defects. These classes are distinguished rates of quantum circuit and TLS, respectively. The by their local symmetries with respect to inversion. strong coupling regime allows one to directly access and Defects which are inversion symmetric about their mid manipulate the TLS quantum state using the circuit as point have (relatively) low bias energy (ε/kB < 10K) a bus, enabling new types of experiments that help to and do not couple to lattice phonons to first order. In understand the TLS origin (see SectionVC). 10

Charge fluctuations - In the first model, TLS are the majority of qubit experiments performed in the last perceived as atomic-sized electric dipoles, which couple decade are consistently explained by assuming that TLS to the oscillating electric fields E in capacitive circuit couple to the Josephson junction’s electric field rather components and tunnel junction barriers. The coupling than modifying its critical current, as explained previ- can be described by ously. However, with the recent realisation that surfaces of quantum devices host a very high density of fluctuating  2 1 (p) magnetic moments [72, 74], the community has largely fo- qˆ − 2 qTLS σz Hcharge = , (12) cussed on fluctuating magnetic fields as the origin of the 2C low frequency noise in quantum electronics. As such the whereq ˆ is the dynamical charge on the circuit capacitor critical current model as an explanation for the observed C and qTLS is the change in induced charge on the ca- 1/f noise has somewhat fallen our of favour. pacitor associated with a change in the state of the TLS. One notable experiment which found direct indica- Equivalently, one can describe this situation as an elec- tion that TLS modify the critical current of a Josephson tric dipole, formed by the TLS, coupled to an electric field junction was reported by Zaretskey et al. [104]. Here, induced by the charges on the circuit capacitor [48, 97]. the spectrum of a Cooper-pair box was observed to be The coupling strength between circuit and TLS is then twinned, displaying multiple parabolas shifted both in given by g = p · E where p is the TLS’ dipole moment frequency and offset in gate charge. These observations and E is the electric field at the TLS position. are consistent with a TLS that couples to both the volt- Charge TLS residing within the tunnel barrier of age across the junctions and the critical current, where Josephson junctions can be exposed to relatively high the latter was found to fluctuate by a large relative value electric fields of up to several hundred V/m and can of 30-40%. As such a result was only reported once so therefore readily be in the strong coupling regime. far, its significance remains however unclear. Charged TLS residing on the circuit substrate or in amor- Finally, magnetic impurities on the surfaces of the phous surface oxides of electrodes will still couple to the quantum circuits may provide fluctuations of the mag- stray electric fields induced by the circuit dynamics, al- netic field threading SQUID loops of the circuits. Such beit more weakly, and are thus thought to lead mostly to loops are typically used in superconducting electronics dielectric loss and energy relaxation [7]. as an effective means to make the Josephson energy EJ Critical current fluctuations - In the second model, tuneable by an applied magnetic field [105]. The coupling the two TLS states are associated with different trans- will be decribed by parencies of a Josephson junction tunnel barrier, corre-  2π  1  sponding to a change of the junction’s critical current. (p) Hmagnetic = Ej cos Φx + ΦTLSσz cosϕ ˆ , Since the rate of Cooper-pair tunnelling across the tunnel Φ0 2 barrier falls off exponentially with distance, at rough in- (14) terfaces the current is transported through a discrete set where Φ is the change in magnetic flux in the SQUID of conductance channels. Thus, blockage of one channel TLS loop due to a change in the state of the TLS, and by e.g., a displaced charge may have a large impact on the Φ is the magnetic flux quantum. For small fluxes total conduction, especially for small area junctions [98– 0 Φ  Φ , the interaction manifests itself as effective 100] and may bring TLS into the strong coupling regime. TLS 0 fluctuations in Josephson energies, very similar to the A similar magnitude of critical current variation might be effect of a fluctuating critical current [106]. Magnetic due to a strongly coupled Kondo impurity in the junction impurities arising from adsorbed surface spins such as dielectric [101, 102]. All these microscopic mechanisms molecular oxygen and atomic hydrogen are thought to be will lead to a coupling between the TLS and a supercon- mainly responsible for the low-frequency magnetic noise, ducting circuit through a modification of the Josephson as their coupling to the circuit dynamics is in general energy in the Hamiltonian, weak [71–74]. However, recent experiments [75] have 1 demonstrated a correlation between removal of surface H = δE σ(p) cosϕ ˆ , (13) current 2 j z spins via annealing and a reduction in the dielectric noise of resonators, suggesting a tantalising link between (p) where σz indicates the state of the TLS (c.f. Eq. (1)), magnetic and charge noise. andϕ ˆ is the superconducting phase difference across the circuit’s Josephson junction. The coupling strength Determining the type of interaction - There are Φ0 δEj = 2π δIc between qubit and TLS is here directly pro- several possibilities of how to distinguish the type of in- portional to the variation δIc of the critical current corre- teraction a given TLS has with their hosting device. In sponding to the two states of the TLS. Here, Φ0 = h/2e general, each type of interaction discussed above leads is the superconducting magnetic flux quantum. to a different term in the Hamiltonian. With enough The critical current coupling was presumed to generate control over individual Hamiltonian parameters, experi- the 1/f noise observed in JJs and SQUIDS [20, 103] and ments can be designed to determine not just the strength also limit the coherence time of qubits through the as- but also the exact form of the interaction, which in turn sociated fluctuations of the qubit energy [18]. However, might allow one to learn the microscopic origin of the 11

TLS under study. Cole et al. [106] compared experi- to its ground state (relaxation), Γ↑ the rate of transi- mental data on two strongly coupled TLS in a phase tions from the ground to the TLS’ excited state, and qubit with a range of theoretical models. Their analy- Γϕ is the TLS’ pure dephasing rate. Here, D[ˆo]ρ = † 1 † †
sis was able to place strong bounds on the parameters of oρˆ oˆ − 2 oˆ oρˆ + ρoˆ oˆ is a Lindblad dissipator, describing some of the microscopic models in the literature, specifi- incoherent processes associated with the operatoro ˆ. The cally ruling out magnetic dipoles and severely restricting strong coupling regime between TLS and its hosting cir- 1 the Andreev level fluctuator hypotheses, but was ulti- cuit is possible if the TLS decoherence rate Γ2 = 2 Γ1+Γϕ mately not able to pin down a specific interaction model. is smaller than its coupling strength g to the circuit. Ad- Zhang and Yu [107] proposed a similar experiment using ditionally, the circuit’s decoherence rate (defined analo- strongly coupled TLS in a flux qubit, as present e.g. in gously) has also to be smaller than g. Here Γ1 = Γ↓ + Γ↑ the experiments of Lupa¸scu et al. [108]. Here the differ- is the inverse TLS lifetime. ent symmetries of the flux qubit Hamiltonian would make The canonical source of dissipation and decoherence a spectroscopy experiment sensitive to different degrees for TLS is a coupling to phonon modes in their hosting of freedom than for the phase qubit used in Cole et al. material [2, 37, 94]. The physical mechanism of this cou- [106], allowing one to overcome the constraints of the ear- pling is the variation in groundstate energy in each well lier experiments and further constricting the microscopic of the TLS due to the variation of the surrounding po- models. tential structure through interactions with phonons and Alternatively, in the case where the coupling between strain, and the subsequent variation in the asymmetry TLS and devices is not limited to a single type of energy ε of the TLS [110]: interaction, one can test for cross-correlations between noise fluctuations in different Hamiltonian parameters ε = 2 γ · S + 2 p · E + ε0. (16) to determine the type of coupling. This method was first proposed for fluctuations in bias charge and critical Here, γ is a tensor defining the TLS’ coupling strength current of the Josephson junction in a phase qubit [109], to the strain field S, the second term accounts for the but so far not implemented in experiments. coupling of the TLS’ electric dipole moment p to the electric field E, and ε0 is an offset imposed by the TLS’ Strength of the interaction - When talking about local environment. For the magnitude of the so-called de- interactions between TLS and quantum circuits, an im- formation potential |γ|, typical values of order ≈ 1 eV are portant concept is again given by the strong coupling found with TLS ensembles from acoustic experiments in regime, i.e. when the strength of interaction g is larger glasses [110], consistent with observations of strain-tuned than the individual decoherence rates of both circuit and individual TLS in the AlOx tunnel barriers of Josephson TLS, g  ΓQ, ΓT LS. In this case, the TLS is coher- junctions [111]. For charged TLS in piezoelectric sub- ently coupled, allowing one to manipulate its state and strates, the additional electric field component E asso- probe its dynamics directly, as will be discussed in more ciated with the lattice vibrations will lead to enhanced detail in Sec.V. In the opposite case of weak coupling, coupling to phonons [112] and thus stronger TLS dissi- g  ΓQ, ΓT LS, the interaction between circuit and TLS pation. can in general be treated perturbatively and the effect The notion that two-level systems may also interact of the TLS will be to provide an effective noise spectral with conduction electrons is based on observations that function to the circuit [20, 40]. This applies equally to TLS in metallic glasses posses enhanced energy relax- individual as well as ensembles of TLS and is most easily ation rates [113]. In quantum circuits, charged TLS lo- probed with resonators (Sec.VI) although also applicable cated in surface oxides of superconducting electrodes can to qubits (Sec.VD) still interact with BCS-quasiparticles originating from in- complete electron pairing. This mechanism is similar to the one proposed to be responsible for qubit dissipation B. Interactions with their dissipative environment at elevated sample temperatures, where quasiparticles that are tunneling across a JJ can absorb energy from Apart from interaction with the dynamics of the host- the qubit [114]. The interaction of quasiparticles with ing device, TLS will almost always show dissipative dy- single TLS in the tunnel barrier of a phase qubit was namics, characterized by incoherent state switching and studied theoretically by Zanker et al. [115] with exper- fluctuations in TLS energy [37]. Generally the time evo- iments performed by Bilmes et al. [116] as described in lution of the TLS density matrix ρ can be described by more detail in Sec.V. a master equation of the Lindblad form ρ˙ = − i [H , ρ] TLS C. TLS-TLS interactions 1 + Γ↓D[σ−]ρ + Γ↑D[σ+]ρ + ΓϕD[σz]ρ , (15) 2 Mutual interaction between TLS may occur by both where HTLS describes the TLS’ free evolution, Γ↓ is elastic and electric dipole coupling when the defect sep- the rate of dissipative transitions from the TLS’ excited aration does not exceed a few nanometres. Although 12 such interactions are neglected in the standard tunnel- a) 10 b) ing model, they have previously been invoked to explain S ω) ( qubit or 0 the linewidth broadening of ultrasonically excited TLS MHz resonator ensembles in glasses [117] and their slow fluctuation dy- −10 TLS 1 namics [118, 119]. 0 100 200 300 energy The first direct observation of two strongly interact- time (minutes) ing and coherent TLS was reported by Lisenfeld et al. P(|e >) g [120] from experiments on phase qubits. Here, strain- 0.1 k T tuning spectroscopy (see Sec.VB) was used to map out B the TLSs’ energy levels, and the results found to be con- 0 TLS 2 sistent with a dipolar interaction between two individual 7.32 7.36 7.4 TLS described by frequency (GHz) 0

(p,1) (p,2) Figure 4. a) Top: Telegraphic switching and drifting of the HTLS−TLS = gTLS σ σ (17) z z resonance frequency of a TLS near 7.36 GHz [133]. Bottom: (p,i) TLS resonances measured by direct microwave spectroscopy where σz describes the state of TLS i in its position ba- at the times indicated by blue and red circles in the top panel. sis. In these experiments, the mutual coupling strength b) Illustration of the spectral diffusion mechanism. The res- gTLS was found to be a substantial fraction of the TLS onance frequency of TLS 1 fluctuates due to its interaction level splitting. Earlier experiments in the same group had with TLS 2, which undergoes random thermal state switch- already shown similar but weaker interactions in other ing since its energy is below kB T . A qubit or resonator close TLS, making it evident that TLS-TLS interactions are to resonance with TLS 1 thus experiences a fluctuating en- not uncommon [111]. vironmental noise spectral density S(ω), affecting its energy relaxation rate and resonance frequency. Several groups have pointed out that allowing for TLS- TLS interactions provides self-consistent distributions for STM parameters that are closer to experimental reality than the canonical ones [26, 91] and can explain recent As a second consequence of this effect, the resonance results on fluctuations in superconducting resonators and frequency of a qubit or microwave resonator may fluctu- qubits [25, 121] and on TLS dephasing under the influ- ate in time when its energy is dispersively shifted by the ence of static strain [122, 123]. coupling to a near-resonant TLS that undergoes spec- Interactions between TLS are thought to be an impor- tral diffusion. This causes qubit dephasing and poses a tant mechanism that gives rise to time-dependent fluctu- significant problem for envisioned superconducting quan- ations of quantum device parameters. Here a high energy tum processors because qubits need to be re-calibrated TLS interacts with one or multiple low-frequency TLS, at regular intervals. Such fluctuations were investigated by Schl¨or et al. [129] in a planar transmon qubit, re- whose excitation energy is below kBT , such that they undergo random thermal transitions (i.e. fluctuators). vealing telegraphic switching of the qubit frequency on a In this case, the resonance frequency of the high-energy time-scale of hours and associated changes in qubit de- TLS may depend on the state of the fluctuator. If a coherence rates. In microwave resonators this mecha- TLS is coupled to one dominant thermal fluctuator, its nism causes resonance frequency fluctuations e.g. in the resonance frequency may display telegraphic fluctuations form of telegraphic noise as observed by Lindstr¨om et al. as shown in Fig.4 a). If more fluctuators are involved, [131] and Burnett et al. [132], as well as excessive phase continuous time-dependent drifts of the TLS resonance noise at low frequencies and temperatures as studied by frequency may occur which is known as spectral diffu- Burnett et al. [121]. sion [118]. 134 used mutual TLS interactions to observe the ran- This mechanism has the consequence that the noise dom state-switching of a TLF by monitoring the time- spectral density which a TLS provides for a qubit or dependent frequency shifts of an interacting TLS, which resonator at a given frequency fluctuates due to the itself was measured using a qubit. This technique allowed time-dependent detuning between them, as illustrated them to investigate the dynamics of TLF switching on in Fig.4 b) [25]. Accordingly, qubits display time- time scales spanning milli-seconds to minutes. dependent fluctuations in their energy relaxation rate as One other case where interactions between TLS observed by Paik et al. [6], Bertet et al. [124], O’Malley are thought to be important is the origin of the et al. [125], and Dial et al. [126] and others, who re- low-frequency magnetic flux noise and specifically its port fluctuations up to ±100% on a time scale of sev- frequency dependence. The frequency dependence of eral hours. Most recently, these effects were observed the low-frequency flux noise is found to be ∼ 1/f α using frequency tunable [127] and fixed frequency trans- extending up to GHz frequencies [44, 135]. Here the mon qubits [128, 129], with the results well explained by exponent α is typically of order one and has been shown the interacting TLS model. In microwave resonators, loss to depend on temperature [65], and its value has a rate fluctuations up to 30% were reported by Megrant strong influence on decoherence times of flux-sensitive et al. [130] and attributed to the same mechanism. qubits [136]. Various models have been suggested to 13

(b) swap reproduce the low-frequency spectrum, most of which excite readout excite π ∆t readout t t involve interacting magnetic moments, with a variety of 0.6 (a) 0.2 types of interactions under investigation [89, 137–140]. 0.3 No clear consensus has been reached so far on the exact 0.1 0 origin of the spectral signatures. 8.0 7.9 (GHz) 8.0 Summary - The fact that TLS interact not only with their hosting devices, but also with each other as well as 7.9 7.9 their own environment is what makes them ultimately detrimental to the operation of superconducting devices.

At the same time the interplay of interactions gives us an frequency (GHz) 7.8 0.2 7.8 0.5 1 opportunity to unambiguously determine the microscopic 0.1 0.5 origin of TLS by combining signatures from several dif- 7.8 (GHz) 8.0 0 0 ferent channels into a single experiment. In the following 0 200 we will review the experimental progress so far towards bias flux (arb. units) time ∆t (ns) this goal. Figure 5. Qubit spectroscopy in (a) the frequency-domain and (b) the time-domain. In (a), the qubit’s excited state population P (|1i) (color-coded) is measured after application V. EXPERIMENTS WITH QUANTUM BITS of a long microwave pulse of varying frequency (insets). Reso- nance with individual TLS (dashed horizontal lines) gives rise A. TLS microwave spectroscopy to split resonance peaks, so-called avoided level crossings. (b) In the so-called “swap-spectroscopy” experiment, the qubit is The development of superconducting qubits has pro- first prepared in its excited state by a microwave π-pulse and then tuned for a time ∆t to a varying probe frequency (see vided significantly enhanced opportunities to investigate sequence in inset of upper panel). While the isolated qubit material defects because they can be used to detect in- shows pure exponential decay due to energy relaxation alone dividual TLS and even allow one to control and observe (blue line), the qubit tuned into resonance with a strongly their quantum state dynamics. The first signatures of coupled TLS displays additional oscillations (red line) which strong interaction between qubits and single defects were reflect the redistribution of energy among the two systems found in microwave spectroscopy experiments. Here, the due to quantum-state swapping [133]. qubit’s excitation energy is varied in a range of a few GHz (typically by an applied magnetic field) and tracked by probing its population in response to application of standard tunneling model which reads long microwave pulses of varying frequency. If the qubit d2N p1 − g2/g2 is tuned into resonance with a strongly coupled TLS, the = σA max , (18) signal changes from a simple Lorentzian to a split peak dEdg 2g due to the lifted degeneracy in the coherently coupled where E is the TLS energy, g is the coupling strength system. See Fig.5 (a) for an example of such mea- between qubit and TLS, gmax is a maximal coupling surements. These characteristic avoided level- or anti- strength determined by the largest observed TLS dipole crossings were revealed in pioneering experiments on su- moment, A is the junction area and σ is the material- perconducting phase qubits performed in the group of specific defect density. Fits of this equation to the in- J.M. Martinis [18], whose observation that the distribu- tegrated number of observed splitting sizes show good tion of anti-crossings changed once a sample was cycled agreement with experiments and provide a robust way to room temperature readily indicated the microscopic to estimate the defect density σ. The STM also agrees origin of the underlying TLS. Soon after, spurious reso- with the measured distribution of coupling strengths g as nances due to TLS were also observed in spectroscopy of verified by Palomaki et al. [142] in a current-biased DC- flux qubits [124, 141] and in the so-called Quantronium, SQUID. For large (≈ 1µm2) Al/AlOx junctions, typical which is a type of charge qubit consisting of a Cooper-pair TLS densities per frequency interval and junction area box that is shunted by a large Josephson junction [42]. are found as σ ≈ 0.4 - 0.5 (GHz µm2)−1 [48, 143, 144]. Martinis et al. [48] also recognized that the major Assuming a typical tunnel barrier thickness of 2-3 nm, source of energy relaxation in first generation phase these measurements correspond to defect densities of qubits was due to TLS-induced dielectric loss occuring σ ≈ 102/(µm3 GHz) with maximal observed dipole mo- in the junction barrier and its surrounding insulation ments of pmax ≈ 6 − 8 Debye. Slightly higher densities 2 layer. This was tested by measurements on microwave of σ ≈ 2.4/(µm GHz) were reported for Al/AlOx junc- resonators and qubits fabricated from different materials tions by Gunnarsson et al. [145], who used a custom pro- (AlOx and SiNx) and confirmed by the relations found cess employing SiNx insulation, and by Hoskinson et al. between qubit decoherence and dielectric loss tangents. [146] who investigated a qubit employing Nb/AlOx/Nb- Moreover, an equation for the density of avoided level trilayer junctions and found σ ≈ 2/(µm2 GHz). For com- crossings observed in spectroscopy was derived from the parison, for silicon nitride Si3N4 which is known to have 14 significantly reduced dielectric loss, Khalil et al. [147] ex- total ten Transmon samples and ten charge qubits and tracted a TLS density of only σ ≈ 0.03/(µm2 GHz) from found avoided level crossings with splitting size exceed- measurements on lumped-element resonators. ing 4 MHz in three of them, roughly estimating the den- Spectroscopy on superconducting qubits can also be sity of strongly coupled TLS to about σ ≈ 4.4/(µm2 used to investigate the type of coupling between a qubit GHz). When comparing this number to previous results and TLS. As discussed in Sec.IV, the coupling can be on phase qubits, note that the smallest resolvable split- longitudinal, where the energy of one system depends on ting size depends on the qubit’s coherence time and thus the state of the other (e.g. if the TLS affects the critical spectroscopic line width. current of a Josephson junction), or transversal, where In charge qubits and single-electron transistors (SETs), the TLS and its host circuit can exchange energy (e.g. a a static electric field can be applied across the Josephson TLS coupled via its electric dipole moment). For exam- junction barrier, which for electrically active TLS tunes ple, Lupa¸scu et al. [108] observed the two-photon transi- their asymmetry energy ε. This could be directly ob- tion to the third excited state of a flux qubit-TLS system served by Kim et al. [152] in a cooper pair box qubit, and found that TLS must be two-level or at least highly where the TLS resonance frequencies were found to de- anharmonic systems which are purely transversally cou- pend linearly on the applied static electric field as ex- pled to the qubit. A similar analysis has been done with pected for asymmetric TLS formed by electric dipoles. In TLS in phase qubits [148, 149]. Spectroscopic data on experiments on an SET, Pourkabirian et al. [63] observed multi-photon transitions of TLS in a phase qubit by Bu- that its effective charge bias was subject to a logarithmic shev et al. [148] were further analysed by Cole et al. [106] drift after a sudden voltage step was applied to the gate. in an effort to verify different microscopic TLS models. This can be interpreted as originating in the slow relax- All of these works can be consistently explained by as- ation of TLS into their new ground state due to inversion suming that TLS couple to the capacitive elements of of their asymmetry energy by the gate voltage step. The the quantum circuits via an electric dipole moment, and same group also studied the temperature dependence of some were even able to place strong bounds on alterna- charge noise in an SET and showed that the environmen- tive models [106]. tal TLS were in stronger thermal contact with the SET Even if avoided level crossings are not observed in qubit electrons than with the phonons in the substrate [153]. spectroscopy, their relatively strong coupling to TLS re- The low-frequency noise in SETs due to TLS was inves- siding in other capacitive circuit components can still be tigated also in earlier work by Zorin et al. [154], who detected by resonant enhancements of the qubit energy focussed on the correlations between fluctuations in two relaxation rate, while the larger number of TLS located adjacent SETs, and found that the responsible fluctuat- in regions of weaker electric fields contribute to a back- ing charges were located either in the substrate or in the ground relaxation rate that is independent or only weakly dielectric covering the circuits. dependent on qubit frequency [20]. Barends et al. [7] found qualitative agreement between such data obtained on so-called Xmon-qubits and Monte Carlo simulations of B. TLS strain-spectroscopy random TLS distributions, in which defects were assumed to occur in a 3-nm thick oxide layer on the aluminium The asymmetry energy ε of a TLS depends linearly electrodes of the coplanar qubit capacitor at similar den- on the local electric field and mechanical strain as given sities as verified for AlOx tunnel barriers. by Eq. (16). The latter effect provides a convenient way When the qubit resonance frequency is being tuned to tune TLS resonance frequencies in a given sample. quickly, Landau-Zener transitions may occur when the To control the mechanical strain, Grabovskij et al. [111] qubit is swept through resonance with strongly coupled used a piezo actuator that slightly bent a chip contain- TLS. This results in a reduction of the readout fidelity ing a phase qubit (see inset of Fig.6), and spectroscopic of phase qubits [150] and in additional losses during measurements confirmed that TLS resonance frequencies qubit operations. indeed depend hyperbolically on the applied strain as expected from the standard tunneling model, see Eq. (5) TLS in charge qubits and single-electron tran- and Eq. (16). From hyperbolic fits to strain-spectroscopy sistors - Experiments with charge qubits, Transmons data, typical values of the TLS’ deformation potential of and flux qubits only rarely observe avoided level cross- γ ≈ 0.1 − 1 eV were extracted, which are consistent with ings. These devices employ very small Josephson junc- measurements on bulk glasses [45]. tions with typical areas of 0.01 − 0.1 µm2, in con- Further strain-tuning experiments by the same group trast to phase qubits in which junctions have sizes of have provided expressive portraits of the TLS distri- 1 − 10 µm2. Moreover, TLS densities in submicron-sized bution as shown in Fig.6[120]. Here, TLS that were junctions were reported to be even lower than expected strain-tuned into resonance with the qubit were detected from the statistical scaling with junction area accord- by their enhancement of the qubit relaxation rate. Such ing to Eq. (18), giving rise to speculation about self- data also reveal mutual TLS interaction in the form annealing effects [15] and the role of reduced film stress of avoided level crossings, non-hyperbolic traces, and in smaller geometries [30]. Schreier et al. [151] tested in telegraphic switching of TLS resonance frequencies. 15

(i) randomly occurring defects considered here, certain kinds of TLS in nearly crystalline materials, which presumably display a high degree of coherence, may still become use- 9 ful for quantum information processing applications. (ii) A phase qubit that is strongly coupled to a TLS displays beating Rabi oscillation when the system is 8 resonantly driven [112]. This was studied as a function of drive amplitude [156] and detuning [157]. In the latter qubit chip work, it was found that a Raman-type transition exists frequency (GHz) in the detuned system which allows one to directly ma- δ P 7 0 Vp nipulate the TLS’ quantum state by resonant microwave driving while the qubit remains in its ground state. This -0.4 piezo technique was used by Lisenfeld et al. [158] to probe 0 100 relative mechanical strain (10-7) the temperature dependence of TLS energy relaxation and dephasing rates, which at elevated temperatures Figure 6. Resonances of TLS (dark traces) in dependence were found to exceed the rates due to TLS-phonon of the mechanical strain applied to a qubit chip. δP (color- coupling. The responsible mechanism was identified coded) indicates the reduction in qubit population due to en- by Bilmes et al. [116] to originate in the interaction ergy absorption from resonant TLS. Mutual TLS interaction of TLS with BCS quasiparticles, where TLS couple to causes telegraphic switching of TLS resonance frequencies (i), the evanescent electronic wave function that leaks from non-hyperbolic traces (ii) and avoided level-crossings (circles). the junction electrodes into the tunnel barrier. This Inset: Illustration of how the mechanical strain was controlled work also showed that one may obtain information by bending the qubit chip with a piezo actuator. about the location of TLS across the tunnel barrier by injecting quasiparticles either into the junction’s top or bottom electrodes, which can provide clues about the fabrication step in which TLS predominantly emerge.

Entanglement can emerge during the time-evolution C. Quantum dynamics of individual TLS of resonantly coupled quantum systems which share a single excitation. The entanglement between the state The strong interaction between a qubit and a TLS can of a phase qubit that was tuned into resonance with be exploited to observe and manipulate the defect’s quan- two TLS and a resonator was observed by Simmonds tum state. When a qubit is prepared in its excited state et al. [143], who found the dynamics of the qubits’ state and tuned into resonance with a TLS, the probability to population to be consistent with the emergence of a find the excitation in the TLS will oscillate at a frequency four-particle entangled system. Grabovskij et al. [159] that corresponds to the qubit-TLS coupling strength. An used a phase qubit to mediate entanglement between example of this so-called quantum state swapping is two TLS by tuning the qubit subsequently into the TLS shown in the upper panel of Figure5 (b). The strong co- resonances and performing a partial swap operation herent coupling to individual TLS is revealed by the char- on each. A similar experiment probed the decay of an acteristic “Chevron”-type pattern as shown in the lower entangled state between a TLS and a resonator [160] panel of Fig.5 (b), displaying the oscillatory redistribu- and again found good agreement with theory. The time tion of energy in the system. By setting the interaction evolution of different entangled states in a resonantly time to half the inverse coupling strength, the quantum coupled qubit-TLS system was observed by Sun et al. states of TLS and qubit are exactly swapped. A TLS can [156], who also studied the emergence of tripartite thus be prepared in an arbitrary quantum state, and like- entanglement via partial Landau-Zener transitions that wise the TLS’ state can be read out by swapping it with occur when an excited phase qubit is swept through the the qubit’s state where it becomes accessible for measure- resonances of two strongly coupled TLS. ment. This technique in principle allows one to use a TLS as a logical qubit as proposed by Zagoskin et al. [155], Measurements of TLS decoherence times - The where the hosting superconducting qubit would merely ability to control and observe the quantum state dy- be used for TLS manipulation and readout. namics of individual TLS provides a way to investigate Resonant swapping of quantum states between a phase the TLS’ interaction with their local environment. By qubit and a TLS has first been observed in the time do- monitoring the dependence of TLS decoherence rates main by Cooper et al. [150]. Subsequently, Neeley et al. on their strain-tuned asymmetry energy, Lisenfeld et al. [19] demonstrated the operation of a TLS as a quantum [122] found evidence that TLS phase coherence is lim- memory by storing the qubit state in the defect and re- ited by their interaction with thermally fluctuating TLS covering it after a waiting time. Although contemporary in their direct vicinity. In addition, strain-independent qubit circuits show much longer coherence times than the maxima observed in TLS’ energy relaxation rates at cer- 16 tain frequencies were attributed to the coupling of TLS to junction electrodes and at edges of interdigitated capac- geometry-specific phonon modes in the Josephson junc- itors, even thin dissipative layers in these regions may tion. have a large impact on the device performance. Swap spectroscopy was used by Shalibo et al. [49] to obtain the statistics of the coupling strengths and coher- A study of qubit energy relaxation rates and their de- ence times of TLS in an AlOx Josephson junction tunnel pendence on the participation ratios of electrode surfaces barrier. They observed TLS lifetimes T1 between 12 ns by Wang et al. [24] found conclusive evidence that di- and 6 µs which were on average anti-correlated with the electric dissipation is a major limiting factor in state-of- coupling strengths to the qubit. Their findings are con- the-art qubits. This result holds both for 3D-Transmon sistent with the scaling of the radiative loss rate due to qubits, which were tested by varying the geometry of ca- phonons with the defect’s dipole size, as expected from pacitor electrodes, as well as for planar Transmons, for the STM. both of which coherence times above 100 µs have been demonstrated [23, 126, 161]. Calculations of the participation ratio can be used D. Dielectric loss and participation ratio to provide information on which region or interface of a quantum circuit contributes most of dielectric loss. When TLS are coupled through an electric dipole mo- Wenner et al. [162] found the substrate-vacuum (S-V) ment to oscillating electric fields within their host de- and metal-substrate (M-S) interfaces to be 100 times vice, they can resonantly absorb energy and give rise to more lossy than the metal-vacuum (M-V) interface dielectric loss. When trying to distinguish and quantify (see Fig.7 for an illustration). This is in accordance losses from different parts of a circuit, a useful concept to Dial et al. [126] and Sandberg et al. [163], who is the participation ratio. It specifies the fraction of the found an order of magnitude smaller participation ratio device’s total energy that is contained in the lossy com- for the M-V interface (p ≈ 0.1 · 10−3) as compared ponent or material. The sum of all losses constitutes to the S-V (p ≈ 1 · 10−3) and the dominating M-S −3 a limit for the circuit’s total energy relaxation time T1, (p ≈ 3 · 10 ) interfaces. The small participation of which can also be described as an internal quality fac- the M-V interface has been attributed to the large mis- tor T1 = Qint/ω with the circuit’s resonance frequency match of dielectric constants [15]. While above studies ω. Employing the concept of the participation ration, we assumed the permittivity of the interfacial layer to be can write this as [24] r ≈ 10, Quintana et al. [164] point out that the relative contribution of the different interfaces depends strongly 1 ω X pi = = ω + Γ . (19) on r. At low r ≈ 2, corresponding to the permittivity T Q Q 0 1 int i i of copolymer resist employed in fabrication, the M-S interface was shown to still participate particularly Here, pi is the participation ratio of the lossy compo- strong, while the M-V and S-V interfaces now contribute nent labelled i, which has an internal quality factor Qi, about equally. Employing this knowledge, it has been and Γ0 is an additional dissipation rate accounting for shown that the influence of TLS in the substrate can be non-dielectric losses. According to the standard tunnel- reduced significantly by etching a trench into the gap ing model, TLS that are themselves interacting with en- region [162, 165–167]. vironmental electric fields and phonons cause dielectric loss rates of [45] Summary - TLS in qubits were first identified as a 1 π|p|2D 1 π|γ|2D major obstacle on the way towards useful superconduct- = 0 , and = 0 , (20) Q 3 Q 2ρv2 ing quantum bits. However the unprecedented degree i,el i i,ph of control that is possible with these circuits, e.g. di- respectively. Here, D0 is the (constant) TLS density of rectly controlling the state of individual TLS and using states, i the permittivity of component material i, ρ is the qubits as probes for their properties, leaves us with the material density, v the sound velocity, and p and γ the very real possibilities to learn all there is to know are the TLS’ electric and elastic dipole moments, respec- about these mysterious defects. However qubit experi- tively. ments and fabrication are particularly challenging, and Since the energy stored in a capacitive component in the following we will review other routes to studying scales with its capacitance Ci and the voltage Ui as these same questions around TLS origin and behaviour. 2 Ei = CiUi /2, the participation ratio increases with the 2 square of the electric field strength i|Ei(r)| integrated over the volume Vi of the lossy component, VI. EXPERIMENTS WITH SUPERCONDUCTING RESONATORS Z  p = i |E (r)|2/E dr , (21) i 2 i tot Vi Superconducting microwave resonators (for a review, with Etot the total electric field energy in the entire see Zmuidzinas [168]) have recently found new appli- space. Due to the high field concentration near tunnel cations as readout devices for superconducting quan- 17 tum bits [27, 28], for qubit interconnections [169, 170], (a) as quantum memories [171], as quantum-limited ampli- ground g w g ground fiers [172], and as detectors for single photons, so-called E kinetic inductance detectors (KIDs) [173]. Resonators are also playing an increasingly important role for the study of TLS in quantum devices, because substrate-vacuum metal-substrate metal-vacuum OH--groups and resist residuals, surface oxides, they can be fabricated with the same technology and other adsorbates, buried adsorbates, adsorbates, materials as quantum bits and are similarly character- process residuals film stress, contaminants ized. Resonators for this purpose are typically shaped cleaning damage C C C C 100 µm into a coplanar configuration as illustrated in Fig.7 a) (b) (c) 3 4 Vbias 2 1 and b), where a central conductor of a certain length is separated by gaps of a few µm size from the surrounding ground planes. Due to the presence of amorphous dielec- tric layers such as surface oxides, these devices will be susceptible to interactions with TLS, which in turn can be used to infer TLS properties. 500µm Alternatively, lumped-element resonators may be used, which comprise discrete planar inductors and capacitors. Figure 7. (a) Sketch of the cross-section through a coplanar These usually employ meandering or coiled-up lines as in- transmission-line resonator, and an overview of mechanisms ductors and interdigitated lines or overlapping films sep- associated with TLS formation at different interfaces (circles). arated by a dielectric layer as capacitors, see Fig.7 c) for (b) Photograph of a typical microwave resonator, having a an example. Lumped-element capacitors bring along the total length of λ/2 ≈ 1 mm at a resonance frequency of 6 GHz, coupled to a transmission line. (c) Lumped-element advantage that the electric field is mostly constrained to resonator which comprises a gradiometric inductor and four the dielectric volume of the capacitors and also homoge- capacitors in a voltage-biased bridge [174]. neous, greatly simplifying the estimation of the participa- tion ratio of TLS-hosting dielectrics. Also, the dielectric in a plate capacitor can be much better defined than the to TLS saturation effects as a function of the circulating spontaneously emerging surface oxides which may be af- power Pint in the resonator is [45, 175] fected by contamination due to air exposure.   Well-known effects that originate in coupling between tanh ~ωR 1 X 2kB T microwave resonators and TLS are power-dependent = pi tan δi r + tan δ0, (22) Qint  β resonator loss, a temperature-dependent resonance fre- i 1 + Pint quency shift, and excessive phase noise due to resonance Pc frequency fluctuations and we summarise each of these where tan δi is the dielectric loss rate due to TLS in vol- in the following. In addition, these effects can be used ume i which has participation ratio p (c.f Eq. (21)), to infer the densities of the TLS involved and resonator i tan δ0 is a residual loss rate due to other mechanisms, ωR structures may allow one to directly manipulate TLS en- is the resonator’s resonance frequency, and T is the tem- sembles with applied electric fields. perature. The sum goes over all lossy components that host TLS, effectively extending Eq. (19) to saturation ef- fects. The exponent β is of order unity and for coplanar A. Power-dependent dielectric loss waveguide resonators is numerically estimated from the geometry to take into account the non-uniformity of the The two-state character of TLS imposes a limit on their electric field distribution. √ contribution to a resonator’s loss rate: once a TLS was The reduction of resonator loss ∝ 1/ Pint in the excited by the resonator field, it has to first dissipate this few photon regime was observed in various experi- energy and return to its ground state before a second pho- ments, e.g. by Lindstr¨om et al. [176], Pappas et al. ton can be absorbed. When the circulating power Pint [177], Ramanayaka et al. [178], and Goetz et al. [179]. in the resonator exceeds a certain critical value Pc, TLS However, measurements on resonators with low intrin- are excited at an effective√ Rabi-frequency that exceeds sic loss rates have shown a much weaker power depen- their loss rates, ΩR > 1/ T1T2. This results in satura- dence than the prediction of Eq. (22)[180–182]. It has tion of the TLS at a stationary excitation probability of been suggested that this effect arises from spectral dif- close to 1/2 with the consequence that the resonator’s loss fusion of strongly interacting TLS at interfaces and on rate is reduced compared to the low-power limit. Here, surface oxides, which causes TLS to drift through the ∆0 ΩR = 2p · E · E /~ is the Rabi frequency for a TLS with resonator’s resonance, effectively suppressing TLS satu- dipole moment p in the resonant electric driving field E, ration [26, 183] and leading to higher loss than predicted with this TLS having an energy relaxation rate 1/T1, de- by Eq. (22). Another possibility is the presence of two phasing rate 1/T2, tunneling energy ∆0, and total energy qualitatively different types of TLS ensembles with dif- E. The STM prediction for the resonator loss rate due ferent critical saturation power Pc as suggested by 91, 18 which would lead to a different saturation behaviour, as such a setup allows one to investigate a broader range of observed e.g. by 184. saturation effects, i.e. the influence of strongly pumping Sage et al. [182] showed that it is possible to actively the TLS in spectral vicinity of one mode on loss and reduce the loss rate of a resonator by applying a strong frequency shift of the second mode. microwave pump tone near the resonance frequency in or- The influence of TLS on a resonators frequency also der to saturate TLS. This method is also known as hole provides a novel pathway towards pinpointing the posi- burning, which refers to the saturation-induced trans- tion of individual TLS and other impurities on surfaces. parency enhancement first observed in materials which Geaney et al. [189] use measurements of frequency shifts are doped with optically active bistable impurities such of a microwave resonator on a quartz tuning fork inte- as dye molecules [2]. Recently similar experiments have grated in a scanning microscope setup to image the sur- probed the change of decay rate and frequency shift of face of a chip containing superconducting metal struc- the resonator when different parts of the TLS ensemble tures. Further improvements in thermal shielding and were saturated, and found good agreement with the pre- isolation of the noise background of the tuning fork res- dictions from the STM for the spectral density of TLS onator are necessary to achieve coherent coupling to in- even at very high frequencies [185]. dividual TLS in that experiment, which will then provide a clear way to identifying individual defects on surfaces, including their exact position and dielectric properties. B. TLS induced resonance frequency shift

The change of the resonance frequency of a resonator C. Noise generated by TLS due to its coupling to a bath of TLS originates in the TLS’ contribution to the dielectric constant , described Besides the so far discussed ensemble effects resulting by [45, 186, 187] in frequency shifts and resonator losses, the coupling to single TLS may also result in a dressing of the resonator 2       ∆ 2D0 p 1 1 hfR hfR states and dispersive resonance shifts. From the Jaynes- = − Re Ψ + − log , Cummings model it follows that in the weak coupling  3 2 2πi kBT kBT (23) regime where the coupling strength g is much smaller than the detuning ∆f between TLS and resonator, g  where Ψ is the complex digamma function and D0 the ∆f, the resonator experiences a dispersive resonance two-level density of states, c.f. Eq. (9). At low tempera- shift ∝ ±g2/∆f depending on the TLS’ state [190]. Ac- tures, the resonant interaction with TLS in their ground cordingly, spectral diffusion of near-resonant TLS can state leads to an increased dielectric constant, while at cause discrete resonance frequency fluctuations, which in higher temperatures incoherent bath induced processes the ensemble limit translate into phase noise as discussed start to dominate and  decreases again. The eigenfre- in the following section. quency of resonators incorporating such dielectrics will Superconducting resonators are usually characterized 1 by measuring the amplitude and phase of a resonant scale accordingly as ∆f/f = − 2 ∆/. In contrast to loss, this frequency shift of the resonator also arises due microwave pulse that is reflected on the resonator. to non-resonant TLS which are not saturated at high While fluctuations of the reflected amplitude concur with power levels, providing a means to characterize the influ- changes in the resonator’s energy relaxation rate, phase ence of TLS on a resonator also with measurements be- noise is related to fluctuations of the resonance frequency. yond the single-photon regime [177]. The non-monotonic Both effects can arise from spectral diffusion of a collec- temperature-dependent frequency shift was studied by tion of near-resonant TLS, which can result in a time- Gao et al. [187] as a function of the center strip width w of dependent spectral density that determines energy re- Nb coplanar resonators (see Fig.7). In resonators where laxation, as well as resonance frequency shifts due to the the electric field was more concentrated (for smaller strip dressing of the resonator transition (see Fig.4). widths and gaps), the frequency shift was more pro- Firm evidence that TLS are a source of resonator phase nounced (∝ 1/w) as expected for TLS that were dis- noise was obtained by Gao et al. [191], who found higher tributed in a few nm-thick oxide layer on the surfaces of noise in resonators that had larger participation ratios superconducting electrodes. Barends et al. [188] showed at lossy interface regions. Further confirmation was pro- that magnitude of the temperature-dependent resonance vided with measurements as a function of the circulat- shift scales with the thickness of a SiO2 layer deposited ing power in the resonator. The noise spectral density −1/2 on top of NbTiN resonators, clearly confirming the role was shown to scale as ∝ Pint , which indicates TLS of TLS in the amorphous capping layer. saturation according to Eq. (22)[176, 186, 188, 190]). Using the coupled symmetric and asymmetric modes These findings are explained by Gao et al. [191] using a of two overlapping coplanar waveguide resonators, 184 semi-empirical model, where the dominant fluctuations investigated the effect of the saturation of the TLSs by are caused by TLS on the electrode surfaces which ex- one mode on the quality factor and frequency shift of perience the strongest electric fields. Similar conclusions the second mode. The availability of multiple modes in were obtained by Neill et al. [192] studying the power- 19

−1 dependence of resonator loss and noise. the spectrum scales as Sy(f) ∝ f (corresponding to −3 Typically, fluctuations of the phase are found to domi- Sϕ(f) ∝ f ), as expected for TLS-induced flicker fre- nate over amplitude noise by as much as 30 dB [190, 193]. quency noise. Takei et al. [194] showed that this effect is due to squeez- By covering resonators with various dielectrics and ing of the noise quadratures by the nonlinearity of TLS observing enhanced noise compared to bare resonators, coupled to the resonator, hereby enhancing the strength several experiments have directly confirmed the role of of phase fluctuations while amplitude noise is suppressed. TLS hosting surface oxides as an origin of resonator For increasing temperature, phase noise typically de- noise [132, 188]. Recently, de Graaf et al. [75] observed creases in amplitude as ∝ T −1−µ in the single pho- a tenfold reduction in the magnitude of frequency fluc- ton regime. Here the exponent µ ranges from 0.2 to tuations in NbN resonators after surface spins such as 0.7 [121, 178], and is associated with the logarithmic physisorbed atomic hydrogen were removed by a ther- temperature dependence of the spectral diffusion width mal annealing treatment. In contrast, losses were re- ∆f(t, T ), i.e. the spectral range over which a TLS dif- duced only weakly by spin desorption. This can be ex- fuses over time [118, 180]. These findings are consis- plained within the frame of the generalized tunneling tent with a generalized tunneling model including inter- model where surface spins take the role of the slowly fluc- actions between high-frequency TLS and thermal fluctu- tuating TLF that generate spectral diffusion of the high- ators [26, 121]. Within this model, at elevated temper- frequency atomic tunneling systems that are responsible atures spectral diffusion of near-resonant TLS plays an for dielectric loss. increasingly minor role because their transitions are al- ready broadened thermally, and the higher decoherence rates of TLS suppresses their interaction with thermal D. TLS density measurements fluctuators. Additionally, it is assumed that TLS at in- terfaces interact more strongly than TLS in the bulk, and The various effects TLS have on microwave resonators that their interaction results in a suppression of the TLS makes these devices useful tools to characterize defect µ/2 density of states ∝ P0E where µ is the same as the densities in deposited materials, and this will continue to exponent in the temperature dependence of the noise. be of importance in the search for low-loss materials for An alternative explanation for these results was given improved solid-state quantum devices. by Burin et al. [195], who argue that mutual TLS inter- For example, Bruno et al. [200] measured the frequency actions are less important at intermediate temperatures dependence of the loss-rate of lumped-element resonators T ≥ 0.1 K and assume that the TLS’ spectral diffusion employing hydrogenated amorphous silicon (a-Si:H) di- width is smaller than their relaxation rate. This latter as- electrics and extract a relatively small loss tangent at sumption was motivated by the early experiment of Bur- 4.2 K of tan δ = 2.5 · 10−5. Smaller TLS densities nett et al. [121] where a Nb resonator was capped with a are observed in a-Si:H due to hydrogen atoms saturat- normal-conducting Pt layer that was expected to enhance ing dangling bonds and increasing the material den- TLS relaxation rates due to their interaction with quasi- sity, which curtails the atomic motional degree of free- particles. However, a later experiment by Burnett et al. dom [201, 202]. In contrast, for thin AlOx layers formed [180] showed consistency with the generalized tunneling by plasma oxidation, Deng et al. [203] obtained the large model also for bare resonators. value tan δ ≈ 1.6·10−3 in agreement to other publications Measurements of the frequency dependence of the discussed here. noise power provide additional clues about the underly- Indications that the TLS density of states increases ing physical mechanism. The standard power-law model monotonically with energy were found by Skacel et al. for noise predicts a scaling of the phase noise power [204], who analysed the frequency-dependent loss in −β spectral density Sϕ(f) ∝ f where the integer val- lumped-element resonators made with amorphous SiO ues β = 0, 1, 2, 3, and 4 are expected for white phase dielectrics. This may stem from mutual TLS interac- noise, flicker phase noise, white frequency noise, flicker tions, which are expected to decrease the density of states frequency noise, and random walk in frequency, respec- n(E) at small energies due to the formation of an Efross- tively [196, 197]. Note that the power spectral den- Shklovskii type pseudogap [26, 121, 183]. However, more sity of the fractional frequency fluctuation Sy(f) is re- measurements and a systematic exploration of materials 2 2 lated to that of phase noise by Sy(f) = (f /f0 )Sϕ(f). is necessary to confirm these findings. Early experiments, where the noise spectrum was probed A robust technique to probe TLS densities in differ- at high circulating powers (with many photons in the ently fabricated Josephson junctions was demonstrated resonator), obtained a frequency dependence close to by Stoutimore et al. [144], who devised a lumped-element −0.5 Sy(f) ∝ f [188, 190, 191, 198]. However, Burnett resonator comprising a meandered inductor, an inter- et al. [132, 180, 199] pointed out that those early re- digitated capacitor, and a Josephson tunnel junction in sults were likely influenced by instrument noise and short similarity to phase qubits. Strong coupling to individ- data acquisition duration, and implemented an improved ual TLS in the junction was measured spectroscopically measurement setup based on a frequency-locked feed- by tuning the resonance frequency via an applied mag- back loop [131] which resulted in clear evidence that netic field and observing avoided level crossings, which 20 occurred at similar densities as found in phase qubit ex- spectral TLS diffusion. periments. A random ensemble of TLS can even be used as a lasing (or strictly speaking masing) medium and coherently amplify the resonator excitation, as was E. Electric field tuning of TLS demonstrated by Rosen et al. [207]. In their experiment, TLS were first inverted into their excited states via TLS that possess an electric dipole moment respond Landau-Zener-transitions by sweeping them electrically to an applied electric DC field by variation of their through resonance with an applied microwave pump asymmetry energy as expressed by Eq. (16). This was tone. Afterwards, the excited TLS were tuned through demonstrated for individual TLS with lumped-element resonance with a resonator, to which they transferred resonators in which the capacitance was formed by their energy by stimulated emission, generating the laser field. four Al/SiNx/Al parallel-plate capacitors in an electri- cal bridge design, allowing one to apply an electric DC- field bias to the dielectric and hereby tune TLS reso- Summary - Resonators are used as tools for precision nance frequencies [174]. Figure7c) shows a photograph measurements in many fields, and the study of TLS in of the design. Using such a device, Khalil et al. [147] amorphous materials turns out to be one of them. Al- observed the dependence of resonator loss on the sweep though they are mostly sensitive to effects from an en- rate of the DC-field. In the strong driving regime, where semble of TLS, rather than individual ones, compared loss is typically reduced due to TLS saturation, enhanced to qubits they allow for more rapid turn-around in ex- loss occurs while the TLS are tuned in frequency by the periment and thus systematic analysis of designs and sweeping the electric field. This is explained by that fact fabrication parameters. In the quest to determine the that TLS which are tuned through the resonator’s transi- microscopic origin of TLS, they are and will remain an tion frequency may absorb energy through Landau-Zener indispensable part of the toolbox. transitions, while their saturation is suppressed due to their short interaction time with the driving field. As demonstrated by Matityahu et al. [205], similar ex- VII. TLS IN OTHER DEVICES periments in a slightly different parameter regime may be effective to decrease dielectric loss from ensembles of There is a variety of other solid-state systems in which TLS. They consider electric field sweeps which periodi- TLS appear to play a role. To highlight the connections cally tune TLS through resonance with a resonator, in the to TLS in quantum circuits, in this section we briefly limit where TLS coherence times are much longer than comment on examples where either the systems in which the sweep period. In this regime, subsequent Landau- the TLS effects are observed displays quantum coherence, Zener transitions of individual TLS typically interfere de- or where the effects of individual TLS can be measured structively such that the TLS’ energy absorption rate is or inferred. effectively reduced, resulting in lower dielectric loss as The coupling of TLS to strain and phonons gives rise verified experimentally. This method can provide a path to a mechanism of damping in nano- and micromechani- to actively decouple the TLS bath from a quantum cir- cal resonators such as suspended beams, cantilevers and cuit. membranes [208–213], and was also shown to affect bulk To extract the distribution of TLS electric dipole mo- acoustic resonators [214–218] and surface-acoustic wave ments p, Sarabi et al. [174] measured the hyperbolic sig- resonators [219]. Here directly the coupling of TLS to natures of individual strongly coupled TLS in the res- phonons is responsible for opening an additional dissipa- onator transmission as a function of frequency and ap- tion channel, similar in spirit to dielectric loss discussed plied bias field. Their experiments showed a broad max- in SectionsVD andVIA. For a more complete overview imum between one to three Debye and extending up to of this field, see Aspelmeyer and Schwab [220] and refer- ∼ 8.3 Debye. Moreover, avoided level crossings were ob- ences therein. served in the transmission when strongly coupled TLS TLS even have a detrimental effect on the quality of were tuned through the resonator’s resonance. By fit- optical devices such as lasers and atomic clocks when ting those to the Jaynes-Cummings model, the dipole they reside in amorphous reflective coatings, where their moments and coherence times of these strongly coupled mechanical fluctuation contributes to thermal noise. This defects can be extracted. Similar results were obtained was reported to be a limiting factor on the finesse of recently by Brehm et al. [206], who investigated TLS re- interferometers used in gravity wave detectors such as siding in an Al/AlOx/Al plate capacitor connected to LIGO [221, 222]. In ion-trap experiments, TLS residing the end of a transmission-line resonator. By tuning TLS on trap electrodes were discussed as a source of so-called via an applied mechanical strain, resonances of individual excess heating and electric-field noise [223]. strongly coupled TLS were detected and their dipole mo- Recently, TLS were reported to be responsible for loss ments and energy relaxation rates were extracted. The in collective electron-spin excitations known as magnons strong resonator-TLS interaction of this system resulted that were observed by coupling a piece of Yttrium-Iron- in pronounced resonator frequency fluctuations due to Garnet (YIG) to a cavity resonator [224, 225]. In 21 this case the coupling between magnon and TLS was A. Removal of dielectrics assumed to occur via phonons. In field effect transis- tors (FETs), the switching of individual two-state defects In conventional microcircuits, deposited dielectrics was identified as the cause of telegraphic noise [226]. The such as amorphous SiO2 are frequently used as insulat- same mechanism gives rise to blinking of fluorescent dye ing layers for the realization of wiring cross-overs and molecules (see Orlov et al. [227] and references therein). on-chip plate capacitors. In quantum circuits, these Defect switching is also thought to be the origin of tele- have to be avoided as they can contribute significantly graphic conductance fluctuations observed in metallic to the total loss. Insulating cross-overs of coplanar res- nanocontacts at intermediate temperatures [228]. In such onators and transmission lines are therefore typically re- a system, the switching rate also varies with the applied alized by so-called airbridges, which are free-standing mechanical strain as expected for atomic tunneling sys- wire interconnects made by depositing a superconducting tems [229]. layer on a photoresist pedestal which is subsequently dis- A recent experiment by Tenorio-Pearl et al. [230] stud- solved [231]. Such airbridges are also required to equal- ied FETs formed by gated 1D nanowires, whose charge ize the ground plane potentials in coplanar resonators distribution was deliberately disordered by defective cap- in order to avoid parasitic slot line modes. Similarly, ping layers of TiO2 or Al2O3. Under application of so-called vacuum gap capacitors have been realized by a microwave drive, the transistor current displayed a reactive ion-etching of the silicon dielectric that sepa- large number of resonances having high quality factors rates two overlapping aluminium electrodes, achieving a ≈ 105, whose resonance frequency changed once the reduction of capacitor loss by one to two orders of mag- sample was thermally cycled. Moreover, the transistor nitude [232, 233]. An often used alternative are planar current showed oscillatory behaviour resembling Larmor capacitors in the form of interdigitated fingers, for which precession and Rabi oscillation in response to pulsed reso- part of the electric field is contained in vacuum. In this nant excitation, with decay times up to several tens of µs. case, coupling to TLS on the substrate and electrode sur- The origin of these resonances was attributed to charged face has to be avoided by limiting the electric field with two-state defects which influence the conductivity of per- an enhanced spacing between electrodes. Experiments by colating current pathways in their vicinity. The observed Sandberg et al. [234] and Gambetta et al. [235] showed resonance frequencies in the GHz range together with the that energy relaxation in contemporary transmon qubits sub-MHz linewidths indicate that these could very well is dominated by capacitor loss when the finger spacing is be the same type of defects, discussed in SectionVC, below 20 to 30 µm. However, strong electric fields can- which are typically observed in superconducting qubits. not be avoided near the qubit’s tunnel junctions. In an effort to reduce substrate loss in this region, Chu et al. [236] etched away the silicon substrate to obtain freely VIII. EMERGENCE OF TLS IN FABRICATION suspended Josephson junctions and DC-SQUIDs. While this treatment resulted in longer qubit T1 times, it also enhanced the level of flux noise, presumably because the It has become increasingly clear that TLS are associ- exposed bottom edge contained a larger density of spins ated with the formation of amorphous interface layers, than the metal-substrate interface of reference samples. disordered materials, and surface adsorbates. In order to avoid and reduce loss from TLS in superconducting cir- cuits, several strategies have been investigated, including: B. Superconducting materials and film deposition • remove lossy dielectrics wherever possible To minimize the density of TLS in amorphous surface layers, devices were fabricated from superconducting ma- • limit the device’s coupling to TLS by employing terials such as rhenium and nitrides including NbN and circuit designs where electric fields are reduced TiN, which are known to develop thinner oxide layers due to their weaker reactivity. The loss rate of resonators fab- • employ less reactive superconducting materials to ricated from epitaxial rhenium was observed to be two to avoid amorphous surface oxides three times lower compared to sputtered aluminum [175]. A study by Sage et al. [182] compared resonators made • carefully optimize fabrication recipes to avoid TLS from poly-crystalline Al, Nb, and TiN films, as well as formation at interfaces epitaxial Al and Rh. They found TiN and Nb to have the lowest and largest losses, respectively. In indepen- • fabricate electrode and tunnel junction barriers dent experiments, NbTiN was shown to exhibit much from crystalline materials. smaller loss [165] and less noise [237] than Nb, Al and Ta. Planar Transmon qubits made with TiN interdigi- Here we summarize some of the main results of recent tated capacitors on nitrided silicon substrates show long studies on the effect of modified fabrication recipes and coherence times up to 60 µs compared to similar devices what they can tell us about origin and location of TLS. made with Al capacitors that achieved ≈ 18µs [238]. 22

However, in practice it is often difficult to attribute D. Substrates losses solely to the materials used, since their deposi- tion and structuring typically involve different techniques Superconducting quantum circuits usually employ sap- and chemistry with corresponding variations of surface phire or high-resistivity silicon substrates. Compared morphology and residuals. For example, sputtered alu- to disordered interfaces, the loss rate of bulk crystalline minium results in rougher films and smaller resonator substrates is several orders of magnitude smaller as was quality factor than Al that is evaporated via electron- shown by Creedon et al. [247],who extracted loss tangents beam or deposited by MBE [130]. Also the growth mode −8 tan δ < 10 for Sapphire (crystalline Al2O3) measur- may have a large impact on losses as was shown by Vis- ing the ring-down of dielectric whispering gallery mode sers et al. [239], who observed about a factor of 10 higher resonators. While the intrinsic loss of silicon substrates quality factors in resonators patterned from mostly (200)- has not been well characterized at low temperatures, [15] TiN polycrystalline films compared to (111)-TiN. A sys- reported that coplanar resonators fabricated on silicon tematic comparison of TiN film properties as a function perform slightly better than those on sapphire. Due to of sputtering parameters we presented by Ohya et al. the minute contribution to the total dissipation of quan- [240], showing that minimal losses of TiN resonators are tum circuits, substrate loss is difficult to identify in ex- associated with reduced film strain and, surprisingly, in- periments. Moreover, direct comparison is complicated creased oxygen content. Moreover, TiN is subject to since different substrates may demand different clean- ageing due to incorporation of contaminants once it is room processing techniques. For example, Gao et al. exposed to air, degrading device performance over time. [190] found smaller levels of phase noise for resonators When fabricating tunnel barriers, Tan et al. [241] produced on sapphire substrates compared to Si or Ge. found that the diffusion process in thermal oxidation of In contrast, Sage et al. [182] reported little influence of Al base electrodes may result in oxygen vacancies which the substrate when comparing different resonator and − bind a layer of chemisorbed O2 . This in turn leads to substrate materials, which may be explained by differ- excess junction noise and larger spread of barrier resis- ent processing steps employed in the former work. No tances over devices. These effects can be mitigated by the difference in resonator loss was found for silicon sub- codeposition of Al and O2 as shown by Welander et al. strates that were capped with either a wet or dry oxide [242]. They obtained ideal subgap resistances of amor- layer, indicating that OH− groups were not a dominant phous AlOx barriers which were codeposited on epitaxial source of TLS in these samples [48]. Recently, Dial et al. Nb/Al base electrodes, while thermal oxidation resulted [126] reported higher quality factors for Transmon qubits in an excess shunt conductance. fabricated on sapphire substrates produced by the heat exchanger method in comparison to the more common edge-defined film-fed grown sapphire. The lack of obvi- ous correlations between such studies of substrate influ- C. Dielectrics ence imply a very strong dependence on the particular fabrication procedure and its associated chemistry. This Similar to superconducting nitrides, nitride dielectrics in turn suggests that systematic (published) studies will such as SiNx are typically less lossy than their oxide coun- be necessary if a general recipe is to be developed which terparts such as SiOx [202]. The loss in amorphous di- can be implemented in a reliable way across different fab- electrics was shown to depend on the material density, in- rication facilities. dicating that the formation of atomic tunneling systems is inhibited in over-constrained materials. For a-Si, this can be achieved by incorporating hydrogen [202, 243] E. Cleaning methods, chemical residuals and film or by deposition at elevated temperatures which leads structuring to more ordered and denser amorphous films [201]. In the latter work, it was shown that growth temperatures To avoid defect formation due to adsorbates and resid- exceeding 350◦C result in TLS-free amorphous films ex- uals, proper cleaning of the substrate prior to material hibiting small loss rates < 2 · 10−7, despite the fact that deposition turns out to be vital. Thermally oxidized sil- they contained no hydrogen but significant dangling bond icon substrates incorporate high densities of coordina- densities as was verified by electron-spin-resonance mea- tion defects at the Si/SiO2 interface [248]. Therefore, surements. cleaning Si substrates in hydrofluoric acid (HF) prior to More recent work has used transmission electron mi- film deposition in order to remove these native oxide and croscopy studies to correlate the structure of amorphous to terminate the surface with hydrogen can significantly aluminium-oxide barriers with low temperature dielec- reduce resonator loss [179, 239]. Megrant et al. [130] tric or tunnelling measurements [244–246], demonstrat- showed that in order to fabricate resonators with quality ing large variability depending on the growth conditions. factors exceeding 106 from Al on sapphire it is crucial In particularly, the influence of the aluminium metal con- that the substrate is first cleaned, e.g. by a reactive oxy- tact morphology on the resulting aluminium-oxide film gen plasma at 850◦C. This is an indication that TLS are quality has been shown to be crucial. formed from adsorbed hydroxyl groups, which are capa- 23 ble of saturating the sapphire (0001) surface [249] and are layer showing a peak in carbon content, followed by a 2 ◦ stable enough to remain even after annealing at 1100 C nm thick layer containing oxygen and AlOx. The latter in UHV [250]. was presumably formed by a reaction of the deposited Al The cleaning method using a reactive oxygen plasma with resist or solvent residues and likely contained high creates less substrate damage than ion-milling, resulting TLS densities. Moreover, it was shown that these resid- in smoother films and a higher quality interface. Quin- uals can be efficiently cleaned using a downstream oxy- tana et al. [164] obtained twice as large resonator quality gen ash descum procedure, which employs neutral oxy- factors and thinner disordered interfacial layers when the gen atoms to chemically remove developed photoresist substrate was cleaned by weak in situ ion-mill (200 eV / from the heated substrate, bringing the losses back to 4 mA for 10 seconds) compared to stronger milling (400 the lower levels found in the etched control resonators. eV / 20 mA for 3.5 minutes). This may indicate that Similar cleaning efficiency was found with an UV-ozone the incorporation of Argon ions or the (disordered) re- process that does not require substrate heating, and is deposition of removed material is associated with TLS therefore preferable for Josephson junction fabrication. formation. Sandberg et al. [163] investigated how different etch- The importance of proper cleaning was further em- ing methods affect the quality of TiN CPW resonators phasized in a study of ageing effects in Josephson junc- on Si substrates, and found that Ar-ion mill patterning tions by Pop et al. [251]. Junction ageing is predomi- caused the formation of amorphous fence-like structures nantly attributed to the presence of chemical contamina- at the electrode etches due to re-deposition of Silicon. tion such as photoresist residues, which thermally diffuse Due to their disordered structure, these fences likely con- into the junction barrier over time. Pop et al. showed tain large densities of TLS which will contribute signif- icantly to the total loss because they are located in re- that completely stable Al/AlOx/Al junctions can be ob- tained if the substrate was initially thoroughly cleaned gions of higher electric field concentration. Lowest loss with an optimized reactive oxygen plasma. This was fur- was achieved using a fluorine-based reactive ion etch, pre- ther cross-checked by annealing junctions in the presence sumably due to its higher edge rates. In contrast to the of a PMMA resist capping layer, which resulted in un- fluorine-based treatment, the chlorine-based process was stable junctions, presumably due to incorporation of hy- suspected to leave Cl salts on the surface and to implant droxides from the resist into the tunnel barrier. This sug- boron ions, leading to more potential contaminants and gests that atomic diffusion of contaminants (either from higher dielectric loss. the substrate preparation or from the resist) may signif- icantly alter the material quality.

The stability of Al/AlOx/Al tunnel junctions can also F. Epitaxial films be improved by annealing finished junctions in a vac- uum chamber at a temperature of 400◦C[252]. More- Soon after it was first recognized that qubit energy re- over, it was shown that annealing also improves the tun- laxation and appearance of avoided level crossings are nelling characteristics overall, as it results in increased associated with TLS in amorphous materials, first at- subgap resistance and the additional disappearance of tempts were made to realize completely crystalline tun- subgap resonances which are due to resonant or inelas- nel junction barriers. Simmonds et al. [255] compared tic tunneling at barrier impurities. It was suggested that the performance of Josephson junctions made with con- the elevated temperatures in the annealing process result ventional ion-mill recipes to those from trilayer processes, in dissociation of aluminium hydrates which may have where the amorphous tunnel barrier is grown in situ on formed during thermal oxidation in the presence of wa- a crystalline Al bottom electrode without breaking vac- ter vapour [253]. The hereby released oxygen is expected uum. While DC transport characteristics such as resid- to combine with the junction electrode material which ual subgap conductance were improved significantly for further increases the tunnel barrier and reduces the crit- crystalline base electrodes, phase qubits made from these ical current, in accordance with the previous results. junctions showed no improvement. However, the high- Outgassing of PMMA photoresist masks during film frequency loss relevant for qubit dissipation may have deposition and their residuals due to incomplete develop- originated from the lossy SiO2 dielectric used for junc- ment is suspected to degrade the quality of Nb resonators tion insulation in all tested samples in that experiment. as reported by Chen et al. [254]. This effect may ex- Patel et al. [256] fabricated crystalline shunt capacitors plain the (by a factor of 3) higher loss rates of resonators by etching a silicon-on-insulator substrate into a mem- patterned in lift-off processes compared to etched res- brane which was covered by Al on both sides. Phase onators as reported by Quintana et al. [164]. In a lift-off qubits employing these shunt capacitors showed T1 times process, the substrate is covered with photoresist prior twice as long as samples with amorphous dielectrics. to metal deposition, and incomplete development may Various problematic effects in the fabrication of epitax- leave residual photoresist at the substrate-metal inter- ial aluminium films were identified by Richardson et al. face. Quintana et al. [164] investigated these residuals [257]. These include corrosion of the Al sidewall due by high-resolution transmission electron microscopy, re- to resist developers, contamination by nanoparticle, and vealing the presence of an 1.6 nm thick resist polymer persistent photoresist residue - all of which could lead 24 to an increased defect density. Josephson junctions with IX. CONCLUSIONS AND OUTLOOK a crystalline Al2O3 tunnel barrier, grown on an epitax- ial Re bottom electrode and capped by polycrystalline In this review, we have summarised the role TLS de- Al, were fabricated by Oh et al. [258]. Phase qubits em- fects play in superconducting circuits, focusing on both ploying these junctions displayed a factor of five smaller the challenges and opportunities that they represent. Al- number of avoided level crossings in spectroscopy, indi- though TLS physics is a relatively old topic of study, cating the better crystallinity of the dielectric. However superconducting circuits have provided an entirely new the qubit’s energy relaxation rate was not reduced com- way of looking at this problem, and an even greater ur- pared to previous samples, presumably due to the pres- gency to solve it. Using superconducting circuits to probe ence of lossy SiO2 wiring insulation in these early exper- TLS provides a range of new possibilities that were sim- iments [259]. ply unavailable to the solid-state glass community over Fully crystalline junctions were fabricated from epi- the previous decades. Probing an individual TLS, de- taxial NbN/AlN/NbN trilayers deposited on MgO (100) termining its characteristic properties including energy substrates by Nakamura et al. [260]. Here special care splitting, decoherence rates, and strain response, all as was taken to grow the AlN tunnel barrier with a cubic a function of applied fields, sample temperature or me- crystal structure in order to avoid piezoelectricity which chanical strain has now become achievable. would lead to conversion of Josephson plasma oscilla- Equally on the side of theory, there are now very spe- tions into phonons. Transmon qubits with these junc- cific targets and goals. The number of relevant mate- tions showed T1 times ranging between 250 and 450 rials is relatively small, the properties of interest are ns. This was significantly longer compared to ∼10 ns very specific and quantitative theory/experiment com- that was achieved with amorphous AlN tunnel barriers parison is becoming the norm. We are now reaching the in early phase qubits [261], but also significantly shorter point where atomistic models can be compared directly than state-of-the-art Transmons which employ shadow- to the (until recently) purely phenomenological parame- evaporated junctions. ters of the standard tunnelling model. Further theoreti- cal advances have the potential to actually predict device Yet another approach was taken by Kline et al. parameters based on fabrication conditions, and signifi- [262], who fabricated junctions with crystalline Sapphire cantly reduce the experimental turn around times. (Al O ) barriers grown on an epitaxial Re/Ti multilayer 2 3 To move forward will require ever closer collaboration base electrode. The multilayer was made by depositing between theory and experiment. New experimental de- 1.5 nm of Ti on 10 nm-thick layers of Re and repeating signs and apparatus that allow measurement of several these steps 12 times. This resulted in much smoother different parameters of a particular TLS need to be de- films (rms roughness of 0.6 nm) as compared to pure Re veloped to be able to estimate these parameters without films which showed terrace structuring and higher rms any unknown scale factors. Honest evaluation of the data roughness of 3.2 nm due to basal-plane twinning [263]. with respect to all relevant theory models is required, However, Junctions that are capped by Re show much without bias towards any particular model. Particularly smaller subgap resistance and poor performance when promising here are efforts that determine the exact type employed in a phase qubit as compared to junctions using of interactions between TLS and their hosting device, Al top electrodes. Weides et al. [264] probed coherence e.g. by exploiting the symmetries of the device Hamil- in Transmon qubits which comprised (ReTi) /Al O /Al 12 2 3 tonian. Knowledge of the type and magnitude of the epitaxial junctions, and measured a small loss tangent of interactions, together with support from theory, will al- 6·10−5 for the crystalline tunnel barrier. Spectroscopy low us to constrain the large zoo of theory models in the on these samples showed a small number of avoided level literature. Similarly, methods that probe the position of crossings due to strongly coupled TLS. However observed individual TLS or TLS ensembles, be it through system- energy relaxation times were still rather short, although atic variation of participation ratios or through applied again this could simply reflect the less mature fabrica- static or spatially dependent fields, have large potential tion and design process compared to existing processes to advance our knowledge of TLS origins when combined and designs. with systematic variations in device fabrication. Finally, Summary - The results discussed in this section a largely unexplored frontier is the modification of the clearly indicate that TLS formation depends sensitively TLS environment, be it through phononic bandgap ma- on the employed materials and techniques of film deposi- terials or quasiparticle generation or trapping. This again tion, structuring, morphology, and substrate treatment. can give us valuable hints towards the microscopic origins It still remains far from clear which combination of fab- of TLS, while at the same time already working towards rication processes promises the greatest improvement in minimizing their impact. coherence and performance of quantum circuits. Nev- Systematic studies of materials, fabrication methods ertheless, it will be crucial that this question is tackled and circuit design dependent performance will provide soon, given the potential benefits of a superconducting new insight and new solutions to this problem. In par- quantum processor and the importance of TLS induced ticular, this should include making publicly available noise and loss for its operation. studies and catalogues of the fabrication processes used, 25 tested and/or discarded. This is particularly important if TABLE OF SYMBOLS the technology is to move from in-house recipes suitable for research publications to large scale device engineer- ing. The experiments needed to advance here are not quick and do not lend themselves to short format, single Table I. Table of all symbols and their definitions. journal article studies. Perseverance is necessary to pros- Symbol Definition per and make progress in this area, but the rewards can ∆0 Tunnelling energy between wells be worthwhile. ε Energy asymmetry between wells

TLS continue to be the Achilles heel of superconduct- E± Eigenenergies of the STM Hamiltonian ing electronics but there is reason for hope. Ironically, d Distance between the wells in the STM using superconducting devices to probe TLS physics pro- V Height of the barrier between the wells vides the very real possibility of solving one of the great P (ε, ∆0) Probability distribution of parameters mysteries of solid-state physics - the true microscopic ori- D(E) Density of states gin of these ubiquitous defects. S(ω) Noise spectral function g Coupling strength C Capacitance

Ej Josephson energy ACKNOWLEDGMENTS Φ0 Magnetic flux quantum Γ Rate of an incoherent process

We thank S. Meissner, M. Schechter, and A. Seiler Γ1 Energy relaxation rate for useful comments and suggestions on the manuscript, Γ2 Total dephasing rate, including relaxation processes and thank J. Burnett for valuable comments on TLS- Γϕ Pure dephasing rate induced phase noise. The authors would also like to ac- γ TLS strain coupling tensor knowledge many useful conversations with the partici- S Strain field tensor pants of the “Tunneling two-level systems and supercon- p Dipole moment vector ducting qubits” workshop held in Sde Boker, September 2016 as well as the workshop “Atomic tunneling Systems E Electric field vector and fluctuating Spins interacting with superconducting σ Defect density Qubits” at MPIPKS in Dresden, February 2019. We Q Quality factor thank Lukas Gr¨unhaupt(Physikalisches Institut, Karl- T1 Energy relaxation time (1/Γ1) sruhe Institute of Technology) for providing Figs.1b) and pi Participation ratio of material component i

1c). JL acknowledges funding from Deutsche Forschungs- i Dielectric constant of material component i gemeinschaft (DFG), project LI2446/1-1. This work tan δ Loss tangent was supported by the Swiss National Science Founda- T Temperature tion through NCCR QSIT and by the Australian Re- P Circulating power in resonator search Council under the Discovery and Centre of Excel- int lence funding schemes (project numbers DP140100375, Pc Critical saturation power CE170100026 and CE110001013).

[1] Christian Enss and Siegfried Hunklinger, Low- Coherence in Josephson Junction Qubits Measured in Temperature Physics (Springer, Berlin, 2005). a Three-Dimensional Circuit QED Architecture,” Phys. [2] P Esquinazi, Tunneling systems in amorphous and crys- Rev. Lett. 107, 240501 (2011). talline solids (Springer, Berlin, 2013). [7] Rami Barends, J Kelly, A Megrant, D Sank, E Jeffrey, [3] John Clarke and Frank K Wilhelm, “Superconducting Yu Chen, Yi Yin, B Chiaro, J Y Mutus, Charles Neill, quantum bits,” Nature 453, 1031–1042 (2008). P O’Malley, Pedram Roushan, James Wenner, T. C. [4] Michel H. Devoret and John M. Martinis, “Implement- White, Andrew N Cleland, and John M Martinis, “Co- ing qubits with superconducting integrated circuits,” herent Josephson qubit suitable for scalable quantum in- Quant. Inform. Process. 3, 163–203 (2004). tegrated circuits,” Phys. Rev. Lett. 111, 080502 (2013). [5] M H Devoret and R J Schoelkopf, “Superconducting [8] Yasunobu Nakamura, Yuri A Pashkin, and Jaw-Shen circuits for quantum information: an outlook.” Science Tsai, “Coherent control of macroscopic quantum states 339, 1169–74 (2013). in a single-Cooper-pair box,” Nature 398, 786–788 [6] Hanhee Paik, D. I. Schuster, Lev S. Bishop, G. Kirch- (1999). mair, G. Catelani, a. P. Sears, B. R. Johnson, M. J. [9] R Barends, J Kelly, A Megrant, A Veitia, D Sank, E Jef- Reagor, L. Frunzio, L. I. Glazman, S. M. Girvin, M. H. frey, T C White, J Mutus, A G Fowler, B Campbell, Devoret, and R. J. Schoelkopf, “Observation of High Y Chen, Z Chen, B Chiaro, A Dunsworth, C Neill, 26

P O’Malley, P Roushan, A Vainsencher, J Wenner, A N ducting circuits,” Phys. Rev.B 92, 035442 (2015). Korotkov, A N Cleland, and John M Martinis, “Super- [26] Lara Faoro and Lev B Ioffe, “Interacting tunneling conducting quantum circuits at the surface code thresh- model for two-level systems in amorphous materials and old for fault tolerance.” Nature 508, 500–3 (2014). its predictions for their dephasing and noise in super- [10] Masoud Mohseni, Peter Read, Hartmut Neven, Sergio conducting microresonators,” Phys. Rev.B 91, 014201 Boixo, Vasil Denchev, Ryan Babbush, Austin Fowler, (2015). Vadim Smelyanskiy, and John Martinis, “Commercial- [27] A. Wallraff, D. I. Schuster, A. Blais, L. Frunzio, R.-S. ize quantum technologies in five years,” Nature , 171 Huang, J. Majer, S. Kumar, S. M. Girvin, and R. J. (2017). Schoelkopf, “Strong coupling of a single photon to a su- [11] M Steffen, M Sandberg, and S Srinivasan, “Recent re- perconducting qubit using circuit quantum electrody- search trends for high coherence quantum circuits,” Su- namics,” Nature 431, 162–167 (2004). percond. Sci. Technol. 30, 030301 (2017). [28] Alexandre Blais, Ren-Shou Huang, Andreas Wallraff, [12] J Niemeyer and V Kose, “Observation of large dc super- Steven M Girvin, and Robert J Schoelkopf, “Cavity currents at nonzero voltages in Josephson tunnel junc- quantum electrodynamics for superconducting electri- tions,” Appl. Phys. Lett. 29, 380–382 (1976). cal circuits: An architecture for quantum computation,” [13] G J Dolan, “Offset masks for lift-off photoprocessing,” Phys. Rev.A 69, 062320 (2004). Appl. Phys. Lett. 31, 337–339 (1977). [29] R. McDermott, “Materials origins of decoherence in su- [14] G J Dolan and J H Dunsmuir, “Very Small (Greater- perconducting qubits,” IEEE Trans. Applied Supercon- Than 20 Nm) Lithographic Wires, Dots, Rings, and ductivity 19, 2–13 (2009). Tunnel-Junctions,” Physica B 152, 7–13 (1988). [30] William D. Oliver and Paul B. Welander, “Materials [15] J. M. Martinis and A. Megrant, “UCSB final report for in superconducting quantum bits,” MRS Bulletin 38, the CSQ program: Review of decoherence and materials 816–825 (2013). physics for superconducting qubits,” arXiv:1410:5793 , [31] F Hund, “Zur Deutung der Molek¨ulspektren. III.” cond–mat/1410.5793 (2014). Zeitschrift f¨urPhysik 43, 805–826 (1927). [16] P Dutta and P M Horn, “Low-Frequency Fluctuations [32] C E Cleeton and N H Williams, “Electromagnetic Waves in Solids - 1-F Noise,” Rev. Mod. Phys. 53, 497–516 of 1.1 cm Wave-Length and the Absorption Spectrum (1981). of Ammonia,” Physical Review 45, 234–237 (1934). [17] E Paladino, Y M Galperin, G Falci, and B L Altshuler, [33] Linus Pauling, “The Rotational Motion of Molecules in “1/f noise: Implications for solid-state quantum infor- Crystals,” Physical Review 36, 430–443 (1930). mation,” Rev. Mod. Phys. 86, 361 (2014). [34] W A Philips, “Tunneling states in amorphous solids,” [18] Raymond W Simmonds, K M Lang, D A Hite, S Nam, J. Low Temp. Phys. 7, 351–360 (1972). David P Pappas, and John M Martinis, “Decoherence [35] P W Anderson, B I Halperin, and C M Varma, in Josephson Qubits from Junction Resonances,” Phys. “Anomalous low-temperature thermal properties of Rev. Lett. 93, 077003 (2004). glasses and spin glasses,” Philosophical Magazine 25, [19] Matthew Neeley, M. Ansmann, Radoslaw C. Bialczak, 1–9 (1972). M. Hofheinz, N. Katz, Erik Lucero, A. O’Connell, [36] W A Philips, Amorphous Solids - Low-Temperature H. Wang, A. N. Cleland, and John M. Martinis, “Pro- Properties, Topics in Current Phys (Springer-Verlag, cess tomography of quantum memory in a Josephson- Berlin, 1981). phase qubit coupled to a two-level state,” Nat. Phys. 4, [37] A W¨urger, From coherent tunneling to relaxation - 523–526 (2008). Dissipative quantum dynamics of interacting defects, [20] Alexander Shnirman, Gerd Sch¨on,Ivar Martin, and Springer tracts in modern physics, Vol. 135 (Springer, Yuriy Makhlin, “Low- and High-Frequency Noise from Berlin, 1997). Coherent Two-Level Systems,” Phys. Rev. Lett. 94, [38] Vassiliy Lubchenko and Peter G Wolynes, “The Mi- 127002 (2005). croscopic Quantum Theory of Low Temperature Amor- [21] Henry J Wold, Hakon Brox, Yuri M Galperin, and phous Solids,” Adv. Chem. Phys. 136, 95–206 (2007). Joakim Bergli, “Decoherence of a qubit due to either a [39] M B Weissman, “1/f noise and other slow, nonexponen- quantum fluctuator, or classical telegraph noise,” Phys. tial kinetics in condensed matter,” Rev. Mod. Phys. 60, Rev.B 86, 205404 (2012). 537–571 (1988). [22] M. Steffen, “Viewpoint: Superconducting Qubits Are [40] Clemens M¨uller, Alexander Shnirman, and Yuriy Getting Serious,” Physics 4, 103 (2011). Makhlin, “Relaxation of Josephson qubits due to strong [23] Chad Rigetti, Jay M. Gambetta, Stefano Poletto, coupling to two-level systems,” Phys. Rev. B. 80, B. L. T. Plourde, Jerry M. Chow, A. D. C´orcoles, 134517 (2009). John A. Smolin, Seth T. Merkel, J. R. Rozen, George A. [41] Kosuke Kakuyanagi, T Meno, Shiro Saito, H Nakano, Keefe, Mary B. Rothwell, Mark B. Ketchen, and Kouichi Semba, Hideaki Takayanagi, Frank Deppe, and M. Steffen, “Superconducting qubit in a waveguide cav- Alexander Shnirman, “Dephasing of a Superconducting ity with a coherence time approaching 0.1 ms,” Phys. Flux Qubit,” Phys. Rev. Lett. 98, 047004 (2007). Rev. B 86, 100506 (2012). [42] G. Ithier, E. Collin, P. Joyez, P. J. Meeson, D. Vion, [24] C. Wang, C. Axline, Y. Y. Gao, T. Brecht, Y. Chu, D. Esteve, F. Chiarello, A. Shnirman, Y. Makhlin, L. Frunzio, M. H. Devoret, and R. J. Schoelkopf, “Sur- J. Schriefl, and G. Sch¨on,“Decoherence in a supercon- face participation and dielectric loss in superconducting ducting quantum bit circuit,” Phys. Rev. B 72, 134519 qubits,” Appl. Phys. Lett. 107, 181103 (2015). (2005). [25] Clemens M¨uller,J¨urgenLisenfeld, Alexander Shnirman, [43] Oleg V Astafiev, Yuri A Pashkin, Yasunobu Nakamura, and Stefano Poletto, “Interacting two-level defects as Tsuyoshi Yamamoto, and Jaw-Shen Tsai, “Quantum sources of fluctuating high-frequency noise in supercon- noise in the Josephson charge qubit,” Phys. Rev. Lett. 27

93, 267007 (2004). [59] Roman M Lutchyn,Lukasz Cywi´nski,Cody P Nave, [44] C M Quintana, Yu Chen, D Sank, A G Petukhov, and S Das Sarma, “Quantum decoherence of a charge T. C. White, Dvir Kafri, B Chiaro, A Megrant, Rami qubit in a spin-fermion model,” Phys. Rev.B 78, 024508 Barends, B Campbell, Z Chen, A Dunsworth, Austin G (2008). Fowler, R Graff, E Jeffrey, J Kelly, Erik Lucero, J Y Mu- [60] Roger H Koch, David P DiVincenzo, and John Clarke, tus, Matthew Neeley, Charles Neill, P O’Malley, Pedram “Model for 1/f flux noise in SQUIDs and qubits,” Phys. Roushan, A Shabani, V N Smelyanskiy, A. Vainsencher, Rev. Lett. 98, 267003 (2007). James Wenner, H Neven, and John M Martinis, “Ob- [61] Kartiek Agarwal, Ivar Martin, Mikhail Lukin, and Eu- servation of Classical-Quantum Crossover of 1 /f Flux gene Demler, “Polaronic model of two-level systems in Noise and Its Paramagnetic Temperature Dependence,” amorphous solids,” Phys. Rev.B 87, 144201 (2013). Phys. Rev. Lett. 118, 057702 (2017). [62] SangKook Choi, Dung-Hai Lee, Steven G Louie, and [45] W. A. Phillips, “Two-level states in glasses,” Reports John Clarke, “Localization of Metal-Induced Gap States on Progress in Physics 50, 1657–1708 (1987). at the Metal-Insulator Interface: Origin of Flux Noise [46] Anthony J Leggett and Dervis C Vural, “”Tunneling in SQUIDs and Superconducting Qubits,” Phys. Rev. two-level systems” model of the low-temperature prop- Lett. 103, 197001 (2009). erties of glasses: are ”smoking-gun” tests possible?”J. [63] A. Pourkabirian, M. V. Gustafsson, G. Johansson, Phys. Chem. B 117, 12966–71 (2013). J. Clarke, and P. Delsing, “Nonequilibrium Probing of [47] Clare C. Yu, “Why study noise due to two level systems: Two-Level Charge Fluctuators Using the Step Response A suggestion for experimentalists,” J. Low Temp. Phys. of a Single-Electron Transistor,” Phys. Rev. Lett. 113, 137, 251–265 (2004). 256801 (2014). [48] J M Martinis, K B Cooper, R McDermott, M Steffen, [64] Roger H Koch, John Clarke, W M Goubau, John M M Ansmann, K D Osborn, K Cicak, S Oh, D P Pap- Martinis, C M Pegrum, and D J Van Harlingen, pas, R W Simmonds, and C C Yu, “Decoherence in “Flicker (1/f) noise in tunnel junction dc SQUIDS,”J. Josephson qubits from dielectric loss,” Phys. Rev. Lett. Low Temp. Phys. 51, 207–224 (1983). 95, 210503 (2005). [65] Frederick C Wellstood, C Urbina, and John Clarke, [49] Y Shalibo, Y Rofe, D Shwa, F Zeides, M Neeley, J M “Low-frequency noise in dc superconducting quantum Martinis, and N Katz, “Lifetime and Coherence of Two- interference devices below 1 K,” Appl. Phys. Lett. 50, Level Defects in a Josephson Junction,” Phys. Rev. Lett. 772 (1987). 105, 177001 (2010). [66] S Sendelbach, D Hover, A Kittel, and M M¨uck, “Mag- [50] Luke Gordon, Hazem Abu-Farsakh, Anderson Jan- netism in SQUIDs at millikelvin temperatures,” Phys. otti, and Chris G Van de Walle, “Hydrogen bonds in Rev. Lett. 100, 227006 (2008). Al2O3 as dissipative two-level systems in superconduct- [67] S. M. Anton, J. S. Birenbaum, S. R. O’Kelley, Vladimir ing qubits,” Scientific Reports 4, 7590 (2014). Bolkhovsky, Danielle A Braje, George Fitch, Matthew [51] Aaron M Holder, Kevin D Osborn, C J Lobb, and Neeley, G. C. Hilton, H. M. Cho, K. D. Irwin, Freder- Charles B Musgrave, “Bulk and surface tunneling hy- ick C Wellstood, William D Oliver, Alexander Shnir- drogen defects in alumina,” Phys. Rev. Lett. 111, man, and John Clarke, “Magnetic Flux Noise in dc 065901 (2013). SQUIDs: Temperature and Geometry Dependence,” [52] Graeme Henkelman and Hannes J´onsson,“Improved Phys. Rev. Lett. 110, 147002 (2013). tangent estimate in the nudged elastic band method for [68] Hendrik Bluhm, Julie A Bert, Nicholas C Koshnick, finding minimum energy paths and saddle points,” J. Martin E Huber, and Kathryn A Moler, “Spin-like sus- Chem. Phys. 113, 9978–9985 (2000). ceptibility of metallic and insulating thin films at low [53] Timothy C DuBois, Manolo C Per, Salvy P Russo, and temperature,” Phys. Rev. Lett. 103, 026805 (2009). Jared H Cole, “Delocalized Oxygen as the Origin of [69] D Lee, J L DuBois, and V Lordi, “Identification of Two-Level Defects in Josephson Junctions,” Phys. Rev. the Local Sources of Paramagnetic Noise in Super- Lett. 110, 077002 (2013). conducting Qubit Devices Fabricated on α-Al2O3 Sub- [54] Timothy C DuBois, Salvy P Russo, and Jared H Cole, strates Using Density-Functional Calculations,” Phys. “Atomic delocalization as a microscopic origin of two- Rev. Lett. 112, 017001 (2014). level defects in Josephson junctions,” New J. Phys. 17, [70] Nicole Adelstein, Donghwa Lee, Jonathan L. DuBois, 023017 (2015). Keith G. Ray, Joel B. Varley, and Vincenzo [55] Timothy C DuBois, Salvy P Russo, and Jared H Cole, Lordi, “Magnetic stability of oxygen defects on “A 3D investigation of delocalised oxygen two-level de- the sio2 surface,” AIP Advances 7, 025110 (2017), fects in Josephson junctions,” arXiv:1508.05204 , cond– https://doi.org/10.1063/1.4977194. mat/1508.05204 (2015). [71] Hui Wang, Chuntai Shi, Jun Hu, Sungho Han, Clare C [56] T C DuBois, M J Cyster, G Opletal, S P Russo, and Yu, and R Q Wu, “Candidate Source of Flux Noise J H Cole, “Constructing ab initio models of ultra-thin in SQUIDs: Adsorbed Oxygen Molecules,” Phys. Rev. Al–AlOx–Al barriers,” Molecular Simulation 42, 542– Lett. 115, 077002 (2015). 548 (2015). [72] P. Kumar, S. Sendelbach, M. A. Beck, J. W. Freeland, [57] J Reinisch and A Heuer, “What is moving in silica at Zhe Wang, Hui Wang, Clare C. Yu, R. Q. Wu, D. P. 1 K? A computer study of the low-temperature anoma- Pappas, and R. McDermott, “Origin and reduction of lies,” Phys. Rev. Lett. 95, 155502 (2005). 1/f magnetic flux noise in superconducting devices,” [58] Alejandro P´erezPaz, Irina V Lebedeva, Ilya V Tokatly, Phys. Rev. Appl. 6, 041001 (2016). and Angel Rubio, “Identification of structural motifs as [73] N. Samkharadze, A. Bruno, P. Scarlino, G. Zheng, tunneling two-level systems in amorphous alumina at D. P. DiVincenzo, L. DiCarlo, and L. M. K. low temperatures,” Phys. Rev.B 90, 224202 (2014). Vandersypen, “High-kinetic-inductance superconduct- 28

ing nanowire resonators for circuit qed in a magnetic Two-Level Systems in Glasses,” Phys. Rev. Lett. 67, field,” Phys. Rev. Appl. 5, 044004 (2016). 2315–2318 (1991). [74] S E de Graaf, A A Adamyan, T Lindstr¨om,D Erts, S E [91] M Schechter and PCE Stamp, “Inversion symmetric Kubatkin, A Ya Tzalenchuk, and A V Danilov, “Direct two-level systems and the low-temperature universality identification of dilute surface spins on Al2O3: Origin of in disordered solids,” Phys. Rev.B 88, 174202 (2013). flux noise in quantum circuits,” Phys. Rev. Lett. 118, [92] Alejandro Gaita-Ari˜noand Moshe Schechter, “Identifi- 057703 (2017). cation of Strong and Weak Interacting Two-Level Sys- [75] S E de Graaf, L Faoro, J Burnett, A A Adamyan, A Ya tems in KBr:CN,” Phys. Rev. Lett. 107, 105504 (2011). Tzalenchuk, S E Kubatkin, T Lindstr¨om, and A V [93] A Churkin, D Barash, and M Schechter, “Nonhomo- Danilov, “Suppression of low-frequency charge noise in geneity of the density of states of tunneling two-level superconducting resonators by surface spin desorption,” systems at low energies,” Phys. Rev. B 89, 104202 Nat. Commun. 9, 1143 (2018). (2014). [76] V G Karpov, M I Klinger, and P N Ignatiev, “Atomic [94] Peter Nalbach and Moshe Schechter, “Symmetry reduc- tunneling states and low-temperature anomalies of ther- tion for tunneling defects due to strong couplings to mal properties in amorphous materials,” Solid State phonons,” New J. Phys. 19, 063030 (2017). Communications 44, 333–337 (1982). [95] M. Schechter, P. Nalbach, and A. L. Burin, “Nonuni- [77] M A Ilin, V G Karpov, and D A Parshin, “Parame- versality and strongly interacting two-level systems in ters of soft atomic potentials in glasses,” JETP 65, 165 glasses at low temperatures,” New J. Phys. 20, 063048 (1987). (2018). [78] U Buchenau, Y M Galperin, V L Gurevich, and [96] A. L. Burin and A. O. Maksymov, “Theory of nonlinear H R Schober, “Anharmonic Potentials and Vibrational microwave absorption by interacting two-level systems,” Localization in Glasses,” Phys. Rev.B 43, 5039–5045 Phys. Rev. B 97, 214208 (2018). (1991). [97] I Martin, L Bulaevskii, and A Shnirman, “Tunnel- [79] F Finkemeier and W von Niessen, “Phonons and phonon ing spectroscopy of two-level systems inside a josephson localization in a-Si: Computational approaches and junction,” Phys. Rev. Lett. 95, 210503 (2005). results for continuous-random-network-derived struc- [98] R. T. Wakai and D. J. Van Harlingen, “Direct lifetime tures,” Phys. Rev.B 58, 4473–4484 (1998). measurements and interactions of charged defect states [80] B B Laird and H R Schober, “Localized Low-Frequency in submicron Josephson junctions,” Phys. Rev. Lett. 58, Vibrational-Modes in a Simple-Model Glass,” Phys. 1687–1690 (1987). Rev. Lett. 66, 636–639 (1991). [99] C D Nugroho, V Orlyanchik, and D J Van Harlingen, [81] B B Laird and S D Bembenek, “Localization and the “Low frequency resistance and critical current fluctu- glass transition,” Journal of Physics-Condensed Matter ations in Al-based Josephson junctions,” Appl. Phys. 8, 9569–9573 (1996). Lett. 102, 142602 (2013). [82] Kostya O Trachenko, Martin T Dove, Mark J Harris, [100] D J Van Harlingen, T L Robertson, B L T Plourde, P A and Volker Heine, “Dynamics of silica glass: two-level Reichardt, T A Crane, and John Clarke, “Decoherence tunnelling states and low-energy floppy modes,” Journal in Josephson-junction qubits due to critical-current fluc- of Physics-Condensed Matter 12, 8041–8064 (2000). tuations,” Phys. Rev.B 70, 064517 (2004). [83] U Zurcher and T Keyes, “Anharmonic potentials in [101] Lara Faoro and Lev B Ioffe, “Microscopic origin of crit- supercooled liquids: The soft-potential model,” Phys. ical current fluctuations in large, small, and ultra-small Rev.E 55, 6917–6927 (1997). area Josephson junctions,” Phys. Rev.B 75, 132505 [84] D C Vural and A J Leggett, “Universal sound absorp- (2007). tion in amorphous solids: A theory of elastically coupled [102] Lara Faoro, Alexei Yu Kitaev, and Lev B Ioffe, “Quasi- generic blocks,” Journal of Non-Crystalline Solids 357, particle poisoning and Josephson current fluctuations 3528–3537 (2011). induced by Kondo impurities,” Phys. Rev. Lett. 101, [85] V Lubchenko, R J Silbey, and P G Wolynes, “Electro- 247002 (2008). dynamics of amorphous media at low temperatures,” [103] Magdalena Constantin and Clare C Yu, “Microscopic Molecular Physics 104, 1325–1335 (2006). Model of Critical Current Noise in Josephson Junc- [86] Lara Faoro, J Bergli, Boris L Altshuler, and Yuri M tions,” Phys. Rev. Lett. 99, 207001 (2007). Galperin, “Models of Environment and T1 Relaxation in [104] V. Zaretskey, B. Suri, S. Novikov, F. C. Wellstood, and Josephson Charge Qubits,” Phys. Rev. Lett. 95, 046805 B. S. Palmer, “Spectroscopy of a cooper-pair box cou- (2005). pled to a two-level system via charge and critical cur- [87] Rogerio de Sousa, K Birgitta Whaley, Theresa Hecht, rent,” Phys. Rev. B 87, 174522 (2013). Jan Von Delft, and Frank K Wilhelm, “Microscopic [105] Yuriy Makhlin, Gerd Sch¨on, and Alexander Shnir- model of critical current noise in Josephson-junction man, “Josephson-Junction Qubits with Controlled Cou- qubits: Subgap resonances and Andreev bound states,” plings,” Nature 398, 305–307 (1998). Phys. Rev.B 80, 094515 (2009). [106] Jared H Cole, Clemens M¨uller, Pavel A Bushev, [88] L Faoro and L B Ioffe, “Quantum two level systems and Grigorij J Grabovskij, J¨urgen Lisenfeld, Alexander kondo-like traps as possible sources of decoherence in Lukashenko, Alexey V Ustinov, and Alexander Shnir- superconducting qubits,” Phys. Rev. Lett. 96, 047001 man, “Quantitative evaluation of defect-models in su- (2006). perconducting phase qubits,” Appl. Phys. Lett. 97, [89] L Faoro and L B Ioffe, “Microscopic origin of low- 252501 (2010). frequency flux noise in Josephson circuits,” Phys. Rev. [107] Zhen-Tao Zhang and Yang Yu, “Coupling mechanism Lett. 100, 227005 (2008). between microscopic two-level system and supercon- [90] S Coppersmith, “Frustrated Interactions and Tunneling: ducting qubits,” Phys. Rev.A 84, 064301 (2011). 29

[108] A. Lupa¸scu,P. Bertet, E. F. C. Driessen, C. J. P. M. and Moshe Schechter, “Decoherence of a quantum two- Harmans, and J. E. Mooij, “One- and two-photon spec- level system by spectral diffusion,” Phys. Rev.B 93, troscopy of a flux qubit coupled to a microscopic defect,” 134208 (2016). Phys. Rev. B 80, 172506 (2009). [124] P. Bertet, I. Chiorescu, G. Burkard, K. Semba, C. J. [109] Lara Faoro and Frank W J Hekking, “Cross correlations P. M. Harmans, D. P. DiVincenzo, and J. E. Mooij, between charge noise and critical-current fluctuations in “Dephasing of a superconducting qubit induced by pho- a four-level tunable Josephson system,” Phys. Rev.B 81, ton noise,” Phys. Rev. Lett. 95, 257002 (2005). 052505 (2010). [125] P. J. J. O’Malley, J. Kelly, R. Barends, B. Camp- [110] J J¨ackle, “On the ultrasonic attenuation in glasses at bell, Y. Chen, Z. Chen, B. Chiaro, A. Dunsworth, low temperature,” Zeitschrift f¨urPhysik 257, 212–223 A. G. Fowler, I.-C. Hoi, E. Jeffrey, A. Megrant, J. Mu- (1972). tus, C. Neill, C. Quintana, P. Roushan, D. Sank, [111] G J Grabovskij, T Peichl, J Lisenfeld, G Weiss, and A. Vainsencher, J. Wenner, T. C. White, A. N. Ko- A V Ustinov, “Strain Tuning of Individual Atomic Tun- rotkov, A. N. Cleland, and John M. Martinis, “Qubit neling Systems Detected by a Superconducting Qubit,” metrology of ultralow phase noise using randomized Science 338, 232–234 (2012). benchmarking,” Phys. Rev. Appl. 3, 044009 (2015). [112] Li-Chung Ku and Clare C. Yu, “Decoherence of a [126] Oliver Dial, Douglas T McClure, Stefano Poletto, G A josephson qubit due to coupling to two-level systems,” Keefe, Mary Beth Rothwell, Jay M Gambetta, David W Phys. Rev. B 72, 024526 (2005). Abraham, Jerry M Chow, and Matthias Steffen, “Bulk [113] J L Black and P Fulde, “Influence of the superconduct- and surface loss in superconducting transmon qubits,” ing state upon the low-temperature properties of metal- Supercond. Sci. Technol. 29, 044001 (2016). lic glasses,” Phys. Rev. Lett. 43, 453 (1979). [127] P V Klimov, J Kelly, Z Chen, M Neeley, A Megrant, [114] Gianluigi Catelani, Robert J Schoelkopf, Michel H De- B Burkett, R Barends, K Arya, B Chiaro, Yu Chen, voret, and Leonid I Glazman, “Relaxation and fre- A Dunsworth, A Fowler, B Foxen, C Gidney, quency shifts induced by quasiparticles in superconduct- M Giustina, R Graff, T Huang, E Jeffrey, Erik ing qubits,” Phys. Rev.B 84, 064517 (2011). Lucero, J Y Mutus, O Naaman, C Neill, C Quin- [115] S Zanker, Michael Marthaler, and Gerd Sch¨on,“De- tana, P Roushan, Daniel Sank, A Vainsencher, J Wen- coherence and Decay of Two-Level Systems Due to ner, T C White, S Boixo, R Babbush, V N Smelyan- Nonequilibrium Quasiparticles,” Ieee Transactions on skiy, H Neven, and John M Martinis, “Fluctuations of Applied Superconductivity 26, 1700204 (2016). Energy-Relaxation Times in Superconducting Qubits,” [116] Alexander Bilmes, Sebastian Zanker, Andreas Heimes, Physical Review Letters 121, 090502 (2018). Michael Marthaler, Gerd Sch¨on,Georg Weiss, Alexey V. [128] J.J Burnett, A. Bengtsson, M. Scigliuzzo, D. Niepce, Ustinov, and J¨urgenLisenfeld, “Electronic decoherence M. Kudra, P. Delsing, and J. Bylander, “Decoherence of two-level systems in a josephson junction,” Phys. Rev. benchmarking of superconducting qubits,” npj Quan- B 96, 064504 (2017). tum Information 5, 54 (2019). [117] W. Arnold and S. Hunklinger, “Experimental evidence [129] S. Schl¨or,J. Lisenfield, C. M¨uller,A. Schneider, D.P. for the direct interaction between two-level systems in Pappas, A.V. Ustinov, and M. Weides, “Correlating glasses at very low temperatures,” Solid State Commun. decoherence in transmon qubits: low frequency noise 17, 883 (1972). by single fluctuators,” arXiv:1901.05352 (2019). [118] J L Black and B I Halperin, “Spectral Diffusion, Phonon [130] A. Megrant, C. Neill, R. Barends, B. Chiaro, Yu Chen, Echoes, and Saturation Recovery in Glasses at Low- L. Feigl, J. Kelly, Erik Lucero, Matteo Mariantoni, Temperatures,” Phys. Rev.B 16, 2879–2895 (1977). P. J. J. O’Malley, D. Sank, A. Vainsencher, J. Wen- [119] A.L. Burin, D. Natelson, D. D. Osheroff, and Y. Kagan, ner, T. C. White, Y. Yin, J. Zhao, C. J. Palmstrøm, “Interactions between tunneling defects in amorphous John M. Martinis, and A. N. Cleland, “Planar su- solids,” in Tunneling Systems in Amorphous and Crys- perconducting resonators with internal quality factors talline Solids, edited by P. Esquinazi (Springer, Berlin, above one million,” Appl. Phys. Lett. 100, 113510 1998) Chap. 5. (2012), https://doi.org/10.1063/1.3693409. [120] J¨urgen Lisenfeld, Grigorij J Grabovskij, Clemens [131] T. Lindstr¨om,J. Burnett, M. Oxborrow, and A Ya. M¨uller,Jared H Cole, Georg Weiss, and Alexey V Usti- Tzalenchuk, “Pound-locking for characterization of su- nov, “Observation of directly interacting coherent two- perconducting microresonators,” Rev. Sci. Instr. 82, level systems in an amorphous material,” Nat. Com- 104706 (2011). mun. 6, 6182 (2015). [132] J. Burnett, T. Lindstr¨om,M. Oxborrow, Y. Harada, [121] Jonathan Burnett, Lara Faoro, I Wisby, V L Gur- Y. Sekine, P. Meeson, and A. Ya. Tzalenchuk, “Slow tovoi, A V Chernykh, G M Mikhailov, V A Tulin, noise processes in superconducting resonators,” Phys. R Shaikhaidarov, V Antonov, P J Meeson, A Ya Tza- Rev. B 87, 140501 (2013). lenchuk, and T Lindstrom, “Evidence for interacting [133] J. Lisenfeld, Data obtained on a superconducting phase two-level systems from the 1/f noise of a superconduct- qubit sample that was fabricated in the group of J. Mar- ing resonator,” Nat. Commun. 5, 4119 (2014). tinis at UCSB, USA, as described in Steffen et al. [265]. [122] J¨urgenLisenfeld, Alexander Bilmes, Shlomi Matityahu, The qubit’s Al/AlOx/Al - Josephson junction had an Sebastian Zanker, Michael Marthaler, Moshe Schechter, area of ≈ 1 µm2. Measurements were done by J. Lisen- Gerd Sch¨on,Alexander Shnirman, Georg Weiss, and feld at Karlsruhe Institute of Technology (KIT), Karl- Alexey V Ustinov, “Decoherence spectroscopy with in- sruhe, Germany. (2011). dividual two-level tunneling defects,” Scientific reports [134] Saskia M. Meißner, Arnold Seiler, J¨urgenLisenfeld, 6, 23786 (2016). Alexey V. Ustinov, and Georg Weiss, “Probing indi- [123] Shlomi Matityahu, Alexander Shnirman, Gerd Sch¨on, vidual tunneling fluctuators with coherently controlled 30

tunneling systems,” Phys. Rev. B 97, 180505 (2018). 82, 134530 (2010). [135] D H Slichter, R Vijay, S. J. Weber, S. Boutin, Maxime [149] Zhican Du, Xueda Wen, Yu Zhou, Guozhu Sun, Jian Boissonneault, Jay M Gambetta, Alexandre Blais, and Chen, and Peiheng Wu, “Spectrum of a superconduct- Irfan Siddiqi, “Measurement-induced qubit state mix- ing phase qubit coupled to a microscopic two-level sys- ing in circuit QED from upconverted dephasing noise,” tem,” Chin. Sci. Bull. 59, 3835–3840 (2014). Phys. Rev. Lett. 109, 153601 (2012). [150] K. B. Cooper, M. Steffen, R. McDermott, R.W. Sim- [136] S. M. Anton, Clemens M¨uller,J. S. Birenbaum, S. R. monds, S. Oh, D. A. Hite, D. P. Pappas, and J.M. O’Kelley, A. D. Fefferman, Dmitri S Golubev, G. C. Martinis, “Observation of quantum oscillations between Hilton, H. M. Cho, K. D. Irwin, Frederick C Wellstood, a Josephson phase qubit and a microscopic resonator us- Gerd Sch¨on,Alexander Shnirman, and John Clarke, ing fast readout.” Phys. Rev. Lett. 93, 180401 (2004). “Pure dephasing in flux qubits due to flux noise with [151] J. A. Schreier, A. A. Houck, Jens Koch, D. I. Schuster, spectral density scaling as 1/f α,” Phys. Rev. B. 85, B. R. Johnson, J. M. Chow, J. M. Gambetta, J. Ma- 224505 (2012). jer, L. Frunzio, M. H. Devoret, S. M. Girvin, and [137] Juan Atalaya, John Clarke, Gerd Sch¨on,and Alexander R. J. Schoelkopf, “Suppressing charge noise decoherence Shnirman, “Flux 1/f α noise in two-dimensional Heisen- in superconducting charge qubits,” Phys. Rev. B 77, berg spin glasses: Effects of weak anisotropic interac- 180502 (2008). tions,” Phys. Rev. B. 90, 014206 (2014). [152] Z. Kim, V. Zaretskey, Y. Yoon, J. F. Schneiderman, [138] K. Kechedzhi, Lara Faoro, and Lev B Ioffe, “Frac- M. D. Shaw, P. M. Echternach, F. C. Wellstood, and tal spin structures as origin of 1/f magnetic noise B. S. Palmer, “Anomalous avoided level crossings in a in superconducting circuits,” arXiv:1102.3445 , cond– Cooper-pair box spectrum,” Phys. Rev. B 78, 144506 mat/1102.3445 (2011). (2008). [139] Herv´eM Carruzzo, Eric R Grannan, and C Yu Clare, [153] M V Gustafsson, A Pourkabirian, G Johansson, “Nonequilibrium dielectric behavior in glasses at low J Clarke, and P Delsing, “Thermal properties of charge temperatures: Evidence for interacting defects,” Phys. noise sources,” Phys. Rev.B 88, 245410 (2013). Rev.B 50, 6685 (1994). [154] A. B. Zorin, F.-J. Ahlers, J. Niemeyer, T. Weimann, [140] Zhi Chen and Clare C Yu, “Comparison of Ising Spin H. Wolf, V. A. Krupenin, and S. V. Lotkhov, “Back- Glass Noise to Flux and Inductance Noise in SQUIDs,” ground charge noise in metallic single-electron tunneling Phys. Rev. Lett. 104, 247204 (2010). devices,” Phys. Rev. B 53, 13682–13687 (1996). [141] B. L. T. Plourde, T. L. Robertson, P. A. Reichardt, [155] A. M. Zagoskin, S. Ashhab, J. R. Johansson, and T. Hime, S. Linzen, C.-E. Wu, and John Clarke, “Flux Franco Nori, “Quantum Two-Level Systems in Joseph- qubits and readout device with two independent flux son Junctions as Naturally Formed Qubits,” Phys. Rev. lines,” Phys. Rev. B 72, 060506 (2005). Lett. 97, 077001 (2006). [142] T. A. Palomaki, S. K. Dutta, R. M. Lewis, A. J. Przy- [156] Guozhu Sun, Zhongyuan Zhou, Bo Mao, Xueda Wen, bysz, Hanhee Paik, B. K. Cooper, H. Kwon, J. R. An- Peiheng Wu, and Siyuan Han, “Entanglement dynam- derson, C. J. Lobb, F. C. Wellstood, and E. Tiesinga, ics of a superconducting phase qubit coupled to a two- “Multilevel spectroscopy of two-level systems coupled level system,” Phys. Rev. B 86, 064502 (2012). to a dc SQUID phase qubit,” Phys. Rev. B 81, 144503 [157] J¨urgen Lisenfeld, Clemens M¨uller, Jared H Cole, (2010). Pavel A Bushev, Alexander Lukashenko, Alexander [143] R. W. Simmonds, M. S. Allman, F. Altomare, K. Ci- Shnirman, and Alexey V Ustinov, “Rabi spectroscopy cak, K. D. Osborn, J. A. Park, M. Sillanp¨a¨a,A. Sirois, of a qubit-fluctuator system,” Phys. Rev. B. 81, 100511 J. A. Strong, and J. D. Whittaker, “Coherent interac- (2010). tions between phase qubits, cavities, and TLS defects,” [158] J¨urgen Lisenfeld, Clemens M¨uller, Jared H Cole, Quant. Inform. Process. 8, 117–131 (2009). Pavel A Bushev, Alexander Lukashenko, Alexander [144] M. J. A. Stoutimore, M. S. Khalil, C. J. Lobb, and K. D. Shnirman, and Alexey V Ustinov, “Measuring the tem- Osborn, “A Josephson junction defect spectrometer for perature dependence of individual two-level systems by measuring two-level systems,” Appl. Phys. Lett. 101, direct coherent control,” Phys. Rev. Lett. 105, 230504 062602 (2012). (2010). [145] David Gunnarsson, Juha-Matti Pirkkalainen, Jian Li, [159] Grigorij J Grabovskij, Pavel A Bushev, Jared H Gheorghe Sorin Paraoanu, Pertti Hakonen, Mika Sil- Cole, Clemens M¨uller, J¨urgen Lisenfeld, Alexander lanp¨a¨a, and Mika Prunnila, “Dielectric losses in multi- Lukashenko, and Alexey V Ustinov, “Entangling mi- layer josephson junction qubits,” Supercond. Sci. Tech- croscopic defects via a macroscopic quantum shuttle,” nol. 26, 085010 (2013). New J. Phys. 13, 063015 (2011). [146] E. Hoskinson, F. Lecocq, N. Didier, a. Fay, F. Hekking, [160] Alexander Kemp, Shiro Saito, William J Munro, Kae W. Guichard, O. Buisson, R. Dolata, B. Mackrodt, and Nemoto, and Kouichi Semba, “Superconducting qubit a. Zorin, “Quantum Dynamics in a Camelback Potential as a quantum transformer routing entanglement be- of a dc SQUID,” Phys. Rev. Lett. 102, 097004 (2009). tween a microscopic quantum memory and a macro- [147] M. S. Khalil, S. Gladchenko, M. J. A. Stoutimore, F. C. scopic resonator,” Phys. Rev.B 84, 104505 (2011). Wellstood, A. L. Burin, and K. D. Osborn, “Landau- [161] J.M. Martinis, “TLS defects and coherence in supercon- Zener population control and dipole measurement of a ducting qubits,” talk at the workshop ”Tunneling Two- two-level-system bath,” Phys. Rev. B 90, 100201 (2014). Level Systems and Superconducting Qubits” given in [148] Pavel A Bushev, Clemens M¨uller, J¨urgen Lisenfeld, Sde Boker, Israel, 2016/11/09. (2016). Jared H Cole, Alexander Lukashenko, Alexander Shnir- [162] J. Wenner, R. Barends, R. C. Bialczak, Yu Chen, man, and Alexey V Ustinov, “Multiphoton spec- J. Kelly, Erik Lucero, Matteo Mariantoni, A. Megrant, troscopy of a hybrid quantum system,” Phys. Rev. B. P. J. J. O’Malley, D. Sank, A. Vainsencher, H. Wang, 31

T. C. White, Y. Yin, J. Zhao, A. N. Cleland, and jected Dipole Moments of Individual Two-Level Defects John M. Martinis, “Surface loss simulations of supercon- Extracted Using Circuit Quantum Electrodynamics,” ducting coplanar waveguide resonators,” Appl. Phys. Phys. Rev. Lett. 116, 167002 (2016). Lett. 99, 113513 (2011). [175] H. Wang, M. Hofheinz, J. Wenner, M. Ansmann, R. C. [163] Martin Sandberg, Michael R. Vissers, Jeffrey S. Kline, Bialczak, M. Lenander, Erik Lucero, M. Neeley, A. D. Martin Weides, Jiansong Gao, David S. Wisbey, and O’Connell, D. Sank, M. Weides, A. N. Cleland, and David P. Pappas, “Etch induced microwave losses in John M. Martinis, “Improving the coherence time of su- titanium nitride superconducting resonators,” Appl. perconducting coplanar resonators,” Appl. Phys. Lett. Phys. Lett. 100, 262605 (2012). 95, 233508 (2009). [164] C. M. Quintana, A. Megrant, Z. Chen, A. Dunsworth, [176] T. Lindstr¨om,J. E. Healey, M. S. Colclough, C. M. B. Chiaro, R. Barends, B. Campbell, Yu Chen, I.-C. Muirhead, and A. Ya. Tzalenchuk, “Properties of su- Hoi, E. Jeffrey, J. Kelly, J. Y. Mutus, P. J. J. O’Malley, perconducting planar resonators at millikelvin temper- C. Neill, P. Roushan, D. Sank, A. Vainsencher, J. Wen- atures,” Phys. Rev. B 80, 132501 (2009). ner, T. C. White, A. N. Cleland, and John M. Martinis, [177] David P. Pappas, Michael R. Vissers, David S. Wisbey, “Characterization and reduction of microfabrication- Jeffrey S. Kline, and Jiansong Gao, “Two Level System induced decoherence in superconducting quantum cir- Loss in Superconducting Microwave Resonators,” IEEE cuits,” Appl. Phys. Lett. 105, 062601 (2014). Transactions on Applied Superconductivity 21, 871–874 [165] R. Barends, N. Vercruyssen, A. Endo, P. J. de Visser, (2011). T. Zijlstra, T. M. Klapwijk, P. Diener, S. J. C. Yates, [178] A. N. Ramanayaka, B. Sarabi, and K. D. Osborn, and J. J. A. Baselmans, “Minimal resonator loss for cir- “Evidence for universal relationship between the mea- cuit quantum electrodynamics,” Appl. Phys. Lett. 97, sured 1/f permittivity noise and loss tangent cre- 023508 (2010). ated by tunneling atoms,” arXiv:1507.06043 , cond– [166] G. Calusine, A. Melville, W. Woods, R. Das, C. Stull, mat/1507.06043 (2015). V. Bolkhovsky, D. Braje, D. Hover, D. K. Kim, [179] Jan Goetz, Frank Deppe, Max Haeberlein, Friedrich X. Miloshi, D. Rosenberg, A. Sevi, J. L. Yoder, Wulschner, Christoph W. Zollitsch, Sebastian Meier, E. Dauler, and W. D. Oliver, “Analysis and mitigation Michael Fischer, Peter Eder, Edwar Xie, Kirill G. Fe- of interface losses in trenched superconducting coplanar dorov, Edwin P. Menzel, Achim Marx, and Rudolf waveguide resonators,” Appl. Phys. Lett. 112, 062601 Gross, “Loss mechanisms in superconducting thin film (2018), https://doi.org/10.1063/1.5006888. microwave resonators,” Journ. Appl. Phys. 119, 015304 [167] A. Bruno, G. De Lange, S. Asaad, K. L. Van Der Enden, (2016). N. K. Langford, and L. Dicarlo, “Reducing intrinsic loss [180] Jonathan Burnett, Lara Faoro, and Tobias Lindstrom, in superconducting resonators by surface treatment and “Analysis of high quality superconducting resonators: deep etching of silicon substrates,” Appl. Phys. Lett. consequences for TLS properties in amorphous oxides,” 106, 182601 (2015). Supercond. Sci. Technol. 29, 044008 (2016). [168] Jonas Zmuidzinas, “Superconducting Microresonators: [181] P. Macha, S. H. W. van der Ploeg, G. Oelsner, Physics and Applications,” Annual Review of Con- E. Il’ichev, H.-G. Meyer, S. W¨unsch, and M. Siegel, densed Matter Physics 3, 169–214 (2012). “Losses in coplanar waveguide resonators at millikelvin [169] Mika A. Sillanp¨a¨a,Jae I. Park, and Raymond W. Sim- temperatures,” Appl. Phys. Lett. 96, 062503 (2010). monds, “Coherent quantum state storage and transfer [182] Jeremy M. Sage, Vladimir Bolkhovsky, William D. between two phase qubits via a resonant cavity,” Nature Oliver, Benjamin Turek, and Paul B. Welander, “Study 449, 438–442 (2007). of loss in superconducting coplanar waveguide res- [170] Matteo Mariantoni, H. Wang, T. Yamamoto, M. Nee- onators,” Journ. Appl. Phys. 109, 063915 (2011). ley, Radoslaw C. Bialczak, Y. Chen, M. Lenander, Erik [183] Lara Faoro and Lev B Ioffe, “Internal Loss of Supercon- Lucero, A. D. O’Connell, D. Sank, M. Weides, J. Wen- ducting Resonators Induced by Interacting Two-Level ner, Y. Yin, J. Zhao, A. N. Korotkov, A. N. Cleland, Systems,” Phys. Rev. Lett. 109, 157005 (2012). and John M. Martinis, “Implementing the Quantum von [184] Naftali Kirsh, E. Svetitsky, Alexander L. Burin, Moshe Neumann Architecture with Superconducting Circuits,” Schechter, and Nadav Katz, “Revealing the nonlinear Science 334, 61–65 (2011). response of a tunneling two-level system ensemble using [171] Max Hofheinz, H. Wang, M. Ansmann, Radoslaw C. coupled modes,” Physical Review Materials 1, 012601 Bialczak, Erik Lucero, M. Neeley, A. D. O’Connell, (2017). D. Sank, J. Wenner, John M. Martinis, and A. N. Cle- [185] Thibault Capelle, E. Flurin, E. Ivanov, J Palomo, land, “Synthesizing arbitrary quantum states in a su- M. Rosticher, S. Chua, T. Briant, P.-F. Cohadon, perconducting resonator,” Nature 459, 546–549 (2009). A. Heidmann, T. Jacqmin, and S. Deleglise, “Prob- [172] N. Bergeal, F. Schackert, M. Metcalfe, R. Vijay, ing a two-level system bath via the frequency shift of an V. E. Manucharyan, L. Frunzio, D. E. Prober, R. J. off-resonantly-driven cavity,” arXiv:1805.04397 (2018). Schoelkopf, S. M. Girvin, and M. H. Devoret, “Phase- [186] Shwetank Kumar, Jiansong Gao, Jonas Zmuidzinas, preserving amplification near the quantum limit with a Benjamin A. Mazin, Henry G. LeDuc, and Pe- Josephson ring modulator,” Nature 465, 64–68 (2010). ter K. Day, “Temperature dependence of the frequency [173] Peter K. Day, Henry G. LeDuc, Benjamin A. Mazin, and noise of superconducting coplanar waveguide res- Anastasios Vayonakis, and Jonas Zmuidzinas, “A onators,” Appl. Phys. Lett. 92, 123503 (2008). broadband superconducting detector suitable for use in [187] Jiansong Gao, Miguel Daal, Anastasios Vayonakis, large arrays,” Nature 425, 817–821 (2003). Shwetank Kumar, Jonas Zmuidzinas, Bernard Sadoulet, [174] Bahman Sarabi, A N Ramanayaka, A L Burin, Fred- Benjamin A. Mazin, Peter K. Day, and Henry G. Leduc, erick C Wellstood, and Kevin D. Osborn, “Pro- “Experimental evidence for a surface distribution of 32

two-level systems in superconducting lithographed mi- 113, 025503 (2014). crowave resonators,” Appl. Phys. Lett. 92, 152505 [202] Aaron D. O’Connell, M. Ansmann, R. C. Bialczak, (2008). M. Hofheinz, N. Katz, Erik Lucero, C. McKenney, [188] R. Barends, H. L. Hortensius, T. Zijlstra, J. J. A. Basel- M. Neeley, H. Wang, E. M. Weig, A. N. Cleland, and mans, S. J. C. Yates, J. R. Gao, and T. M. Klapwijk, J. M. Martinis, “Microwave dielectric loss at single pho- “Contribution of dielectrics to frequency and noise of ton energies and millikelvin temperatures,” Appl. Phys. NbTiN superconducting resonators,” Appl. Phys. Lett. Lett. 92, 112903 (2008). 92, 223502 (2008). [203] Chunqing Deng, M. Otto, and A. Lupascu, “Character- [189] S. Geaney, D. Cox, T. H¨onigl-Decrinis, ization of low-temperature microwave loss of thin alu- R Shaikhaidarov, S E Kubatkin, T Lindstr¨om, minum oxide formed by plasma oxidation,” Appl. Phys. A V Danilov, and S E de Graaf, “Near-field scanning Lett. 104, 043506 (2014). microwave microscopy in the single photon regime,” [204] Sebastian T Skacel, Christoph Kaiser, S Wuensch, arxiv:1902.08066 (2019). Hannes Rotzinger, Alexander Lukashenko, Markus [190] Jiansong Gao, Jonas Zmuidzinas, Benjamin A. Mazin, Jerger, Georg Weiss, Michael Siegel, and Alexey V Usti- Henry G. LeDuc, and Peter K. Day, “Noise proper- nov, “Probing the density of states of two-level tunnel- ties of superconducting coplanar waveguide microwave ing systems in silicon oxide films using superconduct- resonators,” Appl. Phys. Lett. 90, 102507 (2007). ing lumped element resonators,” Appl. Phys. Lett. 106, [191] Jiansong Gao, Miguel Daal, John M. Martinis, Anas- 022603 (2015). tasios Vayonakis, Jonas Zmuidzinas, Bernard Sadoulet, [205] Shlomi Matityahu, H. Schmidt, Alexander Bilmes, Benjamin A. Mazin, Peter K. Day, and Henry G. Leduc, Alexander Shnirman, Georg Weiss, A. V. Ustinov, and “A semiempirical model for two-level system noise in su- J¨urgenLisenfeld, “Dynamical decoupling of quantum perconducting microresonators,” Appl. Phys. Lett. 92, two-level systems by coherent multiple landau-zener 212504 (2008). transitions,” arxiv:1903.07914 (2019). [192] C Neill, A Megrant, R Barends, Yu Chen, B Chiaro, [206] Jan David Brehm, Alexander Bilmes, Georg J Kelly, J Y Mutus, P J J O’Malley, D Sank, J Wenner, Weiss, Alexey V. Ustinov, and J¨urgen Lisenfeld, T C White, Yi Yin, A N Cleland, and John M Marti- “Transmission-line resonators for the study of individ- nis, “Fluctuations from edge defects in superconducting ual two-level tunneling systems,” Appl. Phys. Lett. 111, resonators,” Appl. Phys. Lett. 103, 072601 (2013). 112601 (2017), https://doi.org/10.1063/1.5001920. [193] J. Gao, L. R. Vale, J. A. B. Mates, D. R. Schmidt, [207] Yaniv J. Rosen, Moe S. Khalil, Alexander L. Burin, and G. C. Hilton, K. D. Irwin, F. Mallet, M. A. Castellanos- Kevin D. Osborn, “Random-Defect Laser: Manipulat- Beltran, K. W. Lehnert, J. Zmuidzinas, and H. G. ing Lossy Two-Level Systems to Produce a Circuit with Leduc, “Strongly quadrature-dependent noise in super- Coherent Gain,” Phys. Rev. Lett. 116, 163601 (2016). conducting microresonators measured at the vacuum- [208] Matthias Imboden and Pritiraj Mohanty, “Evidence of noise limit,” Appl. Phys. Lett. 98, 232508 (2011). universality in the dynamical response of micromechan- [194] So Takei, Victor M. Galitski, and Kevin D. Osborn, ical diamond resonators at millikelvin temperatures,” “Squeezed noise due to two-level system defects in su- Phys. Rev. B 79, 125424 (2009). perconducting resonator circuits,” Phys. Rev. B 85, [209] A. Venkatesan, K. J. Lulla, M. J. Patton, A. D. Armour, 104507 (2012). C. J. Mellor, and J. R. Owers-Bradley, “Dissipation due [195] A. L. Burin, M. Schechter, and S. Matityahu, “Low to tunneling two-level systems in gold nanomechanical temperature 1/f noise in microwave dielectric constant resonators,” Phys. Rev. B 81, 073410 (2010). of amorphous dielectrics in josephson qubits,” Phys. [210] F. Hoehne, Yu. A. Pashkin, O. Astafiev, L. Faoro, L. B. Rev. B 92, 174201 (2015). Ioffe, Y. Nakamura, and J. S. Tsai, “Damping in high- [196] Enrico Rubiola, “The Leeson effect - Phase noise in frequency metallic nanomechanical resonators,” Phys. quasilinear oscillators,” arXiv:0502143 (2005). Rev. B 81, 184112 (2010). [197] Enrico Rubiola, Phase noise and frequency stability in [211] J. Suh, A. J. Weinstein, and K. C. Schwab, “Op- oscillators (Cambridge University Press, 2008). tomechanical effects of two-level systems in a back- [198] J. Baselmans, S. J. C. Yates, R. Barends, Y. J. Y. action evading measurement of micro-mechanical mo- Lankwarden, J. R. Gao, H. Hoevers, and T. M. Klap- tion,” Appl. Phys. Lett. 103, 052604 (2013). wijk, “Noise and sensitivity of aluminum kinetic induc- [212] Thomas Faust, Johannes Rieger, Maximilian J. Seitner, tance detectors for sub-mm astronomy,” J. Low Temp. J¨orgP. Kotthaus, and Eva M. Weig, “Signatures of two- Phys. 151, 524–529 (2008). level defects in the temperature-dependent damping of [199] Jonathan Burnett, Andreas Bengtsson, David Niepce, nanomechanical silicon nitride resonators,” Phys. Rev. and Jonas Bylander, “Noise and loss of superconduct- B 89, 100102 (2014). ing aluminium resonators at single photon energies,”J. [213] Matthias Imboden and Pritiraj Mohanty, “Dissipa- Phys. Conf. Ser. 969, 012131 (2018). tion in nanoelectromechanical systems,” Physics Re- [200] A. Bruno, S. T. Skacel, Ch Kaiser, S. W¨unsch, M. Siegel, ports 534, 89 – 146 (2014), dissipation in nano- A. V. Ustinov, and M. P. Lisitskiy, “Investigation of electromechanical systems. dielectric losses in hydrogenated amorphous silicon (a- [214] Guiti Zolfagharkhani, Alexei Gaidarzhy, Seung-Bo Si:H) thin films using superconducting microwave res- Shim, Robert L. Badzey, and Pritiraj Mohanty, “Quan- onators,” in Physics Procedia, Vol. 36 (2012) pp. 245– tum friction in nanomechanical oscillators at millikelvin 249. temperatures,” Phys. Rev. B 72, 224101 (2005). [201] Xiao Liu, Daniel R. Queen, Thomas H. Metcalf, Julie E. [215] C. Seoanez, F. Guinea, and A. H. Castro Neto, “Dis- Karel, and Frances Hellman, “Hydrogen-free amor- sipation due to two-level systems in nano-mechanical phous silicon with no tunneling states,” Phys. Rev. Lett. devices,” EPL (Europhysics Letters) 78, 60002 (2007). 33

[216] Maxim Goryachev, Daniel L. Creedon, Eugene N. White, A. N. Cleland, and John M. Martinis, “Fabri- Ivanov, Serge Galliou, Roger Bourquin, and Michael E. cation and characterization of aluminum airbridges for Tobar, “Extremely low-loss acoustic phonons in a quartz superconducting microwave circuits,” Appl. Phys. Lett. bulk acoustic wave resonator at millikelvin tempera- 104, 052602 (2014). ture,” Appl. Phys. Lett. 100, 243504 (2012). [232] Katarina Cicak, Michael S Allman, Joshua A Strong, [217] R. Rivi`ere,S. Del´eglise,S. Weis, E. Gavartin, O. Ar- Kevin D Osborn, and Raymond W Simmonds, cizet, A. Schliesser, and T. J. Kippenberg, “Optome- “Vacuum-gap capacitors for low-loss superconducting chanical sideband cooling of a micromechanical oscilla- resonant circuits,” IEEE Transactions on Applied Su- tor close to the quantum ground state,” Phys. Rev. A perconductivity 19, 948–952 (2009). 83, 063835 (2011). [233] Katarina Cicak, Dale Li, Joshua A. Strong, Michael S. [218] Kang-Hun Ahn and Pritiraj Mohanty, “Quantum Fric- Allman, Fabio Altomare, Adam J. Sirois, Jed D. tion of Micromechanical Resonators at Low Tempera- Whittaker, John D. Teufel, and Raymond W. Sim- tures,” Phys. Rev. Lett. 90, 085504 (2003). monds, “Low-loss superconducting resonant circuits us- [219] R. Manenti, M. J. Peterer, A. Nersisyan, E. B. Mag- ing vacuum-gap-based microwave components,” Appl. nusson, A. Patterson, and P. J. Leek, “Surface acoustic Phys. Lett. 96, 093502 (2010). wave resonators in the quantum regime,” Phys. Rev. B [234] Martin Sandberg, Michael R. Vissers, Thomas A. 93, 041411 (2016). Ohki, Jiansong Gao, Jos´e Aumentado, Martin Wei- [220] Markus Aspelmeyer and Keith Schwab, “Focus on me- des, and David P. Pappas, “Radiation-suppressed chanical systems at the quantum limit,” New J. Phys. superconducting quantum bit in a planar ge- 10, 095001 (2008). ometry,” Appl. Phys. Lett. 102, 072601 (2013), [221] I W Martin, R Nawrodt, K Craig, C Schwarz, R Bassiri, https://doi.org/10.1063/1.4792698. G Harry, J Hough, S Penn, S Reid, R Robie, and [235] J. M. Gambetta, C. E. Murray, Y. K. K. Fung, D. T. S Rowan, “Low temperature mechanical dissipation of McClure, O. Dial, W. Shanks, J. W. Sleight, and an ion-beam sputtered silica film,” Classical and Quan- M. Steffen, “Investigating surface loss effects in super- tum Gravity 31, 035019 (2014). conducting transmon qubits,” IEEE Transactions on [222] Jonathan P. Trinastic, Rashid Hamdan, Chris Billman, Applied Superconductivity 27, 1700205 (2017). and Hai-Ping Cheng, “Molecular dynamics modeling of [236] Y. Chu, C. Axline, C. Wang, T. Brecht, Y. Y. mechanical loss in amorphous tantala and titania-doped Gao, L. Frunzio, and R. J. Schoelkopf, “Sus- tantala,” Phys. Rev. B 93, 014105 (2016). pending superconducting qubits by silicon micro- [223] M. Brownnutt, M. Kumph, P. Rabl, and R. Blatt, “Ion- machining,” Appl. Phys. Lett. 109, 112601 (2016), trap measurements of electric-field noise near surfaces,” https://doi.org/10.1063/1.4962327. Rev. Mod. Phys. 87, 1419–1482 (2015). [237] R. Barends, N. Vercruyssen, A. Endo, P. J. de Visser, [224] Dany Lachance-Quirion, Yutaka Tabuchi, Arnaud T. Zijlstra, T. M. Klapwijk, and J. J. A. Basel- Gloppe, Koji Usami, and Yasunobu Nakamura, “Hy- mans, “Reduced frequency noise in superconducting res- brid quantum systems based on magnonics,” Applied onators,” Appl. Phys. Lett. 97, 033507 (2010). Physics Express 12, 070101 (2019). [238] Josephine B. Chang, Michael R. Vissers, Antonio D. [225] Marco Pfirrmann, Isabella Boventer, A. Schneider, Tim C´orcoles,Martin Sandberg, Jiansong Gao, David W. Wolz, Mathias Kl¨aui, A. V. Ustinov, and Martin Abraham, Jerry M. Chow, Jay M. Gambetta, Mary Weides, “Magnons at low excitations: Observation of Beth Rothwell, George A. Keefe, Matthias Steffen, and incoherent coupling to a bath of two-level-systems,” David P. Pappas, “Improved superconducting qubit co- arXiv:1903.03981 (2019). herence using titanium nitride,” Appl. Phys. Lett. 103, [226] M.J. Kirton and M.J. Uren, “Noise in solid-state mi- 012602 (2013). crostructures: A new perspective on individual defects, [239] M. R. Vissers, J. Gao, D. S. Wisbey, D. A. Hite, C. C. interface states and low-frequency 1/f noise,” Advances Tsuei, A. D. Corcoles, M. Steffen, and D. P. Pap- in Physics 38, 367–468 (1989). pas, “Low loss superconducting titanium nitride copla- [227] S. V. Orlov, A. V. Naumov, Yu. G. Vainer, and nar waveguide resonators,” Appl. Phys. Lett. 97, 232509 Lothar Kador, “Spectrally resolved analysis of fluores- (2010). cence blinking of single dye molecules in polymers at [240] S Ohya, B Chiaro, A Megrant, C Neill, R Barends, low temperatures,” J. Chem. Phys. 137, 194903 (2012). Y Chen, J Kelly, D Low, J Mutus, P J J O’Malley, [228] K. S. Ralls and R. A. Buhrman, “Defect Interactions P Roushan, D Sank, A Vainsencher, J Wenner, T C and Noise in Metallic Nanoconstrictions,” Phys. Rev. White, Y Yin, B D Schultz, C J Palmstrøm, B A Lett. 60, 2434–2437 (1988). Mazin, A N Cleland, and John M Martinis, “Room [229] S. Brou¨er,G. Weiss, and H. B. Weber, “Mechanically temperature deposition of sputtered tin films for super- controlled tunneling of a single atomic defect,” Euro- conducting coplanar waveguide resonators,” Supercond. phys. Lett. 54, 654 (2001). Sci. Technol. 27, 015009 (2014). [230] J.O. Tenorio-Pearl, E.D. Herbschleb, S. Fleming, [241] E. Tan, P. G. Mather, A. C. Perrella, J. C. Read, and C. Creatore, S. Oda, W.I. Milne, and A.W. Chin, R. A. Buhrman, “Oxygen stoichiometry and instability “Observation and coherent control of interface-induced in aluminum oxide tunnel barrier layers,” Phys. Rev. B electronic resonances in a field-effect transistor,” Nature 71, 161401 (2005). Materials 16, 208 (2017). [242] Paul B. Welander, Timothy J. McArdle, and James N. [231] Zijun Chen, A. Megrant, J. Kelly, R. Barends, Eckstein, “Reduced leakage current in josephson tunnel J. Bochmann, Yu Chen, B. Chiaro, A. Dunsworth, junctions with codeposited barriers,” Appl. Phys. Lett. E. Jeffrey, J. Y. Mutus, P. J. J. O’Malley, C. Neill, 97, 233510 (2010). P. Roushan, D. Sank, A. Vainsencher, J. Wenner, T. C. [243] Robert O. Pohl, Xiao Liu, and EunJoo Thompson, 34

“Low-temperature thermal conductivity and acoustic cond. Sci. Technol. 21, 075013 (2008). attenuation in amorphous solids,” Rev. Mod. Phys. 74, [255] R.W Simmonds, D.A. Hite, R. McDermott, M. Stef- 991–1013 (2002). fen, K.B. Cooper, K.M. Lang, J.M. Martinis, and [244] L J Zeng, P Krantz, S Nik, P Delsing, and E Ols- D.P. Pappas, “Junction materials research using phase son, “The atomic details of the interfacial interaction qubits,” in Quantum Computing in Solid State Systems between the bottom electrode of Al/AlOx/Al Joseph- (Springer, 2004) Chap. 11. son junctions and HF-treated Si substrates,” Journal of [256] U Patel, Y Gao, D Hover, G J Ribeill, S Sendelbach, Applied Physics 117, 163915 (2015). and R McDermott, “Coherent Josephson phase qubit [245] Lunjie Zeng, Dung Trung Tran, Cheuk-Wai Tai, Gunnar with a single crystal silicon capacitor,” Appl. Phys. Lett. Svensson, and Eva Olsson, “Atomic structure and oxy- 102, 012602 (2013). gen deficiency of the ultrathin aluminium oxide barrier [257] C J K Richardson, N P Siwak, J Hackley, Z K Keane, in Al/AlOx/Al Josephson junctions,” Scientific Reports J E Robinson, B Arey, I Arslan, and B S Palmer, “Fab- 6, 29679 (2016). rication artifacts and parallel loss channels in metamor- [246] S Fritz, A Seiler, L Radtke, R Schneider, M Weides, phic epitaxial aluminum superconducting resonators,” G Weiss, and D Gerthsen, “Correlating the nanostruc- Supercond. Sci. Technol. 29, 064003 (2016). ture of Al-oxide with deposition conditions and dielec- [258] S Oh, K Cicak, R McDermott, K B Cooper, K D tric contributions of two-level systems in perspective of Osborn, R W Simmonds, M Steffen, J M Martinis, superconducting quantum circuits,” Scientific Reports and D P Pappas, “Low-leakage superconducting tunnel 8, 9 (2018). junctions with a single-crystal Al2O3 barrier,” Super- [247] Daniel L. Creedon, Yarema Reshitnyk, Warrick Farr, cond. Sci. Technol. 18, 1396 (2005). John M. Martinis, Timothy L. Duty, and Michael E. [259] Seongshik Oh, Katarina Cicak, Jeffrey S Kline, Mika A Tobar, “High q-factor sapphire whispering gallery mode Sillanp¨a¨a, Kevin D Osborn, Jed D Whittaker, Ray- microwave resonator at single photon energies and mil- mond W Simmonds, and David P Pappas, “Elimination likelvin temperatures,” Appl. Phys. Lett. 98, 222903 of two level fluctuators in superconducting quantum bits (2011). by an epitaxial tunnel barrier,” Phys. Rev.B 74, 100502 [248] C R Helms and E H Poindexter, “The silicon-silicon (2006). dioxide system: Its microstructure and imperfections,” [260] Y. Nakamura, H. Terai, K. Inomata, T. Yamamoto, Reports on Progress in Physics 57, 791 (1994). W. Qiu, and Z. Wang, “Superconducting qubits con- [249] J Ahn and J. W. Rabalais, “Composition and structure sisting of epitaxially grown NbN/AlN/NbN josephson of the Al2O3(0001)(1×1) Surface,” Surf. Sci. 388, 121– junctions,” Appl. Phys. Lett. 99, 212502 (2011). 131 (1997). [261] John M. Martinis, “Superconducting phase qubits,” [250] C. Niu, K. Shepherd, D. Martini, J. Tong, J.A. Kel- Quant. Inform. Process. 8, 81–103 (2009). ber, D.R. Jennison, and A. Bogicevic, “Cu interactions [262] Jeffrey S Kline, Michael R Vissers, Fabio C S da Silva, with α-Al2O3(0001): effects of surface hydroxyl groups David S Wisbey, Martin Weides, Terence J Weir, Ben- versus dehydroxylation by ar-ion sputtering,” Surf. Sci. jamin Turek, Danielle A Braje, William D Oliver, Yoni 465, 163 – 176 (2000). Shalibo, Nadav Katz, Blake R Johnson, Thomas A [251] I. M. Pop, T. Fournier, T. Crozes, F. Lecocq, I. Matei, Ohki, and David P Pappas, “Sub-micrometer epitaxial B. Pannetier, O. Buisson, and W. Guichard, “Fabrica- josephson junctions for quantum circuits,” Supercond. tion of stable and reproducible submicron tunnel junc- Sci. Technol. 25, 025005 (2012). tions,” J. Vac. Sci. Tech. B 30, 010607 (2012). [263] Paul B. Welander, “Structural evolution of Re (0001) [252] P. J. Koppinen, L. M. V¨aist¨o,and I. J. Maasilta, “Com- thin films grown on Nb (110) surfaces by molecular plete stabilization and improvement of the characteris- beam epitaxy,” Journ. Appl. Phys. 108, 103508 (2010). tics of tunnel junctions by thermal annealing,” Appl. [264] Martin P Weides, Jeffrey S Kline, Michael R Vissers, Phys. Lett. 90, 053503 (2007). Martin O Sandberg, David S Wisbey, Blake R Johnson, [253] J. V. Gates, M. A. Washington, and M. Gurvitch, Thomas A Ohki, and David P Pappas, “Coherence in a “Critical current uniformity and stability of transmon qubit with epitaxial tunnel junctions,” Appl. nb/aloxidenb josephson junctions,” Journ. Appl. Phys. Lett. 99, 262502 (2011). Phys. 55, 1419–1421 (1984). [265] Matthias Steffen, M. Ansmann, R. McDermott, [254] Wei Chen, Douglas A Bennett, Vijay Patel, and N. Katz, Radoslaw Bialczak, Erik Lucero, Matthew James E Lukens, “Substrate and process dependent Neeley, E. Weig, a. Cleland, and John Martinis, “State losses in superconducting thin film resonators,” Super- Tomography of Capacitively Shunted Phase Qubits with High Fidelity,” Phys. Rev. Lett. 97, 050502 (2006).